A Year Just Happened in a Week

That Was The Week

0:00

-9:05

A Year Just Happened in a Week

The Great Leap Forward

Keith Teare

May 24, 2025

Transcript

Editorial: A Year Happened in a Week

This editorial was written by Google’s Notebook LM after reading this week’s collection and being prompted by myself. It is in the spirit of my views as a result. I am traveling and so grabbed the chance to try it out. And as Andrew and I are both traveling the Podcast is also done by Google Notebook LM.

Welcome to "That Was The Week". While each week brings significant developments in the tech and startup world, the past seven days have felt less like a single week and more like an entire year of technological advancement compressed into an intense period. What we've witnessed is nothing short of a Great Leap Forward in artificial intelligence, reshaping industries, devices, and even the fundamental nature of work and competition.

The focus is squarely on AI, demonstrating an extraordinary acceleration not just in capability, but in its practical integration across applications and devices. We are seeing a shift from AI as a conceptual tool to AI as a usable, deeply embedded assistant. This isn't about simple chat interfaces anymore; it's about sustained, autonomous collaboration with users, capable of complex reasoning, coding, multi-modal inputs, and integrating various tools. Examples from the week highlight this, with developments from players like Anthropic demonstrating sustained coding sessions and achieving new benchmarks, and Google introducing enhanced reasoning modes for complex problem-solving. These developments lay the foundation for redefining productivity and user experience.

Alongside the leap in AI models, a crucial new battleground is emerging beyond software: hardware. A major strategic pivot into "physical AI embodiments" is signaled by large acquisitions aimed at building new AI companion devices. These devices are designed potentially to transcend conventional screens, reduce screen dependence, and even challenge entrenched players in the device market. While another major player continues to embed AI heavily within its existing software and services like Search, NotebookLM mobile, and AI Overviews, moves into physical design leadership signify that the competition is increasingly moving into the physical realm, creating new dynamics around device innovation and consumer ownership.

The ripple effects of this AI acceleration are evident across the landscape. Venture capital funding continues to flow heavily into AI, capturing roughly one-third of global VC investment, exceeding $100 billion in 2024. While Series B rounds are showing some volatility, there's a broader pivot towards efficiency and profitability. Simultaneously, seed-stage investing is adapting with new playbooks supporting founders early without necessarily dominating ownership.

Perhaps most immediately impactful for many is the transformation already underway in the workplace. Leading companies are not just recommending AI adoption; they are mandating widespread proficiency. Executives are issuing stark warnings that mastering AI tools is becoming essential, linking job security to demonstrating how tasks can be done more efficiently with AI, and cautioning against professional irrelevance for those who don't adapt. This trend underscores the profound sociological and managerial upheaval AI is causing, influencing workflows, morale, and corporate culture fundamentally.

Beneath these rapid shifts lies a critical, perhaps defining, debate over AI's ultimate architecture. There's a powerful advocacy for an open, protocol-based AI ecosystem fostering interoperability and innovation—what's been called "an architecture of participation". Proponents argue that participatory markets allow solutions to emerge from anywhere, preventing monopolistic control. Yet, the aggressive maneuvers of dominant players building controlling ecosystems and platforms suggest a strong pursuit of winner-takes-most opportunities. This sets the stage for an ideological and practical contest with huge technological, economic, and ethical implications.

This week crystallized a pivotal inflection point—AI is maturing rapidly, its reach is expanding dramatically, and strategic battles are being drawn across software, hardware, and ecosystem control. As we reflect on this "Great Leap Forward", crucial questions remain: Are we witnessing the dawn of truly universal AI assistants integrated seamlessly into our lives, or are we seeing the birth of new digital gatekeepers controlling access and innovation? Navigating this rapid transition to stay ahead will be the defining challenge for individuals and organizations alike in this AI-powered future. What’s clear is this: the year truly just happened, compressed within a single week, and AI stands at the stage center, shaping what comes next.

Essays

24 Years After ‘Sorry, Steve: Here’s Why the Apple Stores Won’t Work’

Ritholtz • John Gruber • May 22, 2025

Technology•Retail•AppleStores•Innovation•MarketingStrategy•Essays

Barry Ritholtz, in an excerpt from his brand-new book, How Not to Invest, marking the occasion of the 24th anniversary of Cliff Edwards’s claim chowder hall of famer, predicting doom for Apple’s then-new foray into its own chain of retail stores:

There are many genuinely revolutionary products and services that, when they come along, change everything. Pick your favorite: the iPod and iPhone, Tesla Model S, Netflix streaming, Amazon Prime, AI, perhaps even Bitcoin. Radical products break the mold; their difference and unfamiliarity challenge us. We (mostly) cannot foretell the impact of true innovation. Then, once it’s a wild success, we have a hard time recalling how life was before that product existed.

The Apple Store was clearly one of those game-changers: By 2020, Apple had opened over 500 stores in 25 countries. They are among the top-tier retailers and the fastest to reach a billion dollars in annual sales. They achieved the highest sales per square foot in 2012 among all retailers. By 2017, they were generating $5,546 per square foot in revenues, twice the dollar amount of Tiffany’s, their closest competitor. Apple no longer breaks out the specifics of its stores in its quarterly reports, but estimates of store revenue are about $2.4 billion per month.

May 2001 is so long ago, Daring Fireball hadn’t yet launched. So I can’t say I predicted the success of Apple’s retail stores. But what I recall thinking, at the time, was that it might work, and was definitely worth trying. Here’s the nut of Edwards’s 2001 piece:

Since PC retailing gross margins are normally 10% or less, Apple would have to sell $12 million a year per store to pay for the space. Gateway does about $8 million annually at each of its Country Stores. Then there’s the cost of construction, hiring experienced staff. “I give them two years before they’re turning out the lights on a very painful and expensive mistake,” says Goldstein. [...]

What’s more, Apple’s retail thrust could be one step forward, two steps back in terms of getting Macs in front of customers. Since most Mac fans already know where to buy, much of the sales from Apple’s stores could come out of the hides of existing Mac dealers. That would bring its already damaged relations with partners to new lows. In early 1999, Best Buy Co. dropped the iMac line after refusing a Jobs edict that it stock all eight colors. Sears, Roebuck & Co. late last year dumped Apple, sources say, after concluding that sales were too hit or miss. And in recent weeks, Mac-only chains such as The Computer Store and ComputerWare have closed down, citing weak margins. Now, faced with competition from Apple, others may cut back. “When you choose to compete with your retailers, clearly that’s not a comfortable situation,” says CompUSA Chief Operating Officer Lawrence N. Mondry.

Two decades later, talking about the importance of Sears as a retail partner looks pretty dumb. But to me, the obvious problem with this argument in 2001 is that if Apple’s existing retail partners in 2001 were going an even vaguely good job, why was the Mac’s market share so low? At the time they were only a handful of years past the crisis where the company almost went bankrupt. Apple, in the old days, had some fantastic small mom-and-pop official retailers, but they were small. And the big partners, like CompUSA, absolutely sucked at showcasing the Mac. Their demo machines frequently broken. If you understood and believed that the Mac was a superior product, it was easy to conclude that its relatively low market share must have been a function of problem with its marketing and retail strategy.

(I’m a longtime fan of Ritholtz’s writing; I’ve got a copy of How Not to Invest and it’s next on my reading list after I finish Patrick McGee’s Apple in China.)

Read More

The Great Product Reset: Will AI Leave Designers Behind?

Newinternet • Jeff Morris Jr. • May 22, 2025

Technology•AI•ProductDesign•CareerDevelopment•Innovation•Essays

The Great Product Reset: Will AI Leave Designers Behind?

I spoke yesterday with one of the most talented product designers I've ever collaborated with, and something he said stuck with me: "I feel old."

He's around 37, an age when professionals in most fields are entering the prime of their careers.

But today he's witnessing the AI-driven product design landscape reinvent itself in a language he hasn't yet learned.

He had a few job interviews recently where he didn't get the role, and I could see the experience woke him up.

So he asked me for advice.

Should he target Series A startups? Update his resume? What could he do this summer?

My answer was pretty simple: Start designing again. But use all the new tools.

Pick a model on Hugging Face. Experiment with GPT-4, Claude, or Gemini.

Or take an app from 2015 and reimagine them with an AI-native products lens.

Evernote. Yelp. Early Instagram. Whatever you want, but start experimenting again.

If you feel the world accelerating past you, you can start doing things.

That's why like Jony Ive joined OpenAI, and why Sergey Brin jumped back in, saying:

"This is the most fun I've had in my life, honestly, and this is the greatest transformative moment in computer science ever. Being a computer scientist, it is the most exciting thing of my life technologically."

Because they love the game. They know the game just reset. And they're endlessly curious, eager to discover what they can build.

One great product idea is all it takes to change everything, to make you relevant again.

That's the unique part of this moment, when building a powerful app is nearly as easy as writing this essay.

Start building today. Not tomorrow, not next week. Today.

That's your way back to relevance and the only path to rediscover your creative confidence.

Now is your moment to fall in love with technology again.

Read More

An Architecture of Participation for AI?

Oreilly • Tim O’Reilly • May 19, 2025

Technology•AI•Innovation•ArchitectureOfParticipation•AIprotocols•Essays

An Architecture of Participation for AI?

About six weeks ago, I sent an email to Satya Nadella complaining about the monolithic winner-takes-all architecture that Silicon Valley seems to envision for AI, contrasting it with “the architecture of participation” that had driven previous technology revolutions, most notably the internet and open source software. I suspected that Satya might be sympathetic because of past conversations we’d had when his book Hit Refresh was published in 2017.

I made the case that we need an architecture for the AI industry that enables cooperating AIs, that isn’t a winner-takes-all market, and that doesn’t make existing companies in every industry simply the colonial domains of extractive AI conquerors, which seems to be the Silicon Valley vision.

Little did I know that Microsoft already had something in the works that is a demonstration of what I am hoping for. It’s called NLWeb (Natural Language Web), and it’s being announced today. Satya offered O’Reilly the chance to be part of the rollout, and we jumped at it.

My ideas are rooted in a notion about how technology markets evolve. We have lived through three eras in computing. Each began with distributed innovation, went through a period of fierce competition, and ended with monopolistic gatekeepers. In the first age (mainframes), it was IBM, in the second (PCs) Microsoft, and in the third (internet and mobile) the oligopoly of Google, Amazon, Meta, and Apple.

The mistake that everyone makes is a rush to crown the new monopolist at the start of what is essentially a wide-open field at the beginning of a new disruptive market. And they envision that monopoly largely as a replacement for what went before, rather than realizing that the paradigm has changed. When the personal computer challenged IBM’s hardware-based monopoly, companies raced to become the dominant personal computer hardware company. Microsoft won because it realized that software, not hardware, was the new source of competitive advantage.

The story repeated itself at the beginning of the internet era. Marc Andreessen’s Netscape sought to replace Microsoft as a dominant software platform, except for the internet rather than the PC. AOL realized that content and community, not software, was going to be a source of competitive advantage on the internet, but they made the same mistake of assuming the end game of consolidated monopoly rather than embracing the early stage of distributed innovation.

So here we are at the beginning of the fourth age, the age of AI, and once again, everyone is rushing to crown the new king. So much of the chatter is whether OpenAI or one of its rivals will be the next Google, when it looks to me that they are more likely the next Netscape or the next AOL. DeepSeek has thrown a bomb into the coronation parade, but we haven’t yet fully realized the depth of the reset, or conceptualized what comes next. That is typically figured out through a period of distributed innovation.

The term “the architecture of participation” originally came to me as an explanation of why Unix had succeeded as a collaborative project despite its proprietary license while other projects failed despite having open source licenses. Unix was designed as a small operating system kernel supporting layers of utilities and applications that could come from anyone, as long as they followed the same rules. Complex behaviors could be assembled by passing information between small programs using standard data formats. It was a protocol-centric view of how complex software systems should be built, and how they could evolve collaboratively. Linux, of course, began as a re-implementation of Unix, and it was the architecture of participation that it inherited, as much as the license and the community, that was the foundation of its success. The internet was also developed as a distributed, protocol-based system.

That concept ran through my web advocacy in the early ’90s, open source advocacy in the late ’90s, and Web 2.0 in the aughts. Participatory markets are innovative markets; prematurely consolidated markets, not so much. The barriers to entry in the early PC market were very low, entrepreneurship high. Ditto for the Web, ditto for open source software and for Web 2.0. For late Silicon Valley, fixated on premature monopolization via “blitzscaling” (think Uber, Lyft, and WeWork as examples, and now OpenAI and Anthropic), not so much. It’s become a kind of central planning. A small cadre of deep-pocketed investors pick the winners early on and try to drown out competition with massive amounts of capital rather than allowing the experimentation and competition that allows for the discovery of true product-market fit.

And I don’t think we have that product-market fit for AI yet. Product-market fit isn’t just getting lots of users. It’s also finding business models that pay the costs of those services, and that create value for more than the centralized platform. The problem with premature consolidation is that it narrows the focus to the business model of the platform, often at the expense of its ecosystem of developers.

As Bill Gates famously told Chamath Palihapitiya when he was running the nascent (and ultimately failed) Facebook developer platform, “This isn’t a platform. A platform is when the economic value of everybody that uses it exceeds the value of the company that creates it. Then it’s a platform.” To be clear, that is not just value to end users. It’s value to developers and entrepreneurs. And that means the opportunity to profit from their innovations, not to have that value immediately harvested by a dominant gatekeeper.

Now of course, Sam Altman talks about creating value for developers. In a recent appearance at Sequoia Capital’s AI Ascent event, he said his hope is to create “like just an unbelievable amount of wealth creation in the world and other people to build on that.” But he uses the language of “an operating system” that others build on top of (and pay OpenAI for the use of) rather than a shared infrastructure co-created by an ecosystem of developers.

That’s why I’ve been rooting for something different. A world where specialized content providers can build AI interfaces to their own content rather than having it sucked up by AI model builders who offer up services based on it to their own users. A world where application developers can offer new kinds of services that enable others in a cooperative cascade.

Anthropic’s Model Context Protocol, an open standard for connecting AI agents and assistants to data sources, is the first step toward a protocol-centric vision of cooperating AIs. It has generated a lot of well-deserved enthusiasm. Google’s A2A takes that further with a vision of how AI agents might cooperate. NLWeb adds to that an easy way for internet content sites to join the party, offering both a conversational front end to their content and an MCP server so that it is accessible to agents.

This is all going to take years to get right. But because it’s a protocol-centric rather than a platform-centric vision, solutions can come from everywhere, not just from a dominant monopolist.

Every new wave of computing has also had a new user interface paradigm. In the mainframe era, it was the teletype terminal; for the PC, the Graphical User Interface; for the internet, the web’s document-centric interface; for mobile, touch screens. For AI (for now at least), it appears to be conversational interfaces.

Companies such as Salesforce and Bret Taylor’s Sierra are betting on conversational agents that are front ends to companies, their services, and their business processes, in the same way that their website or mobile app is today. Others are betting on client-side agents that will access remote sites, but often by calling APIs or even performing the equivalent of screen scraping. MCP, A2A, and other agent protocols point to a richer interaction layer made up of cooperating AIs, able to connect to any site offering AI services, not just via API calls to a dominant AI platform.

All companies need at least a start on an AI frontend today. There’s a fabulous line from C. S. Lewis’s novel Till We Have Faces: “We cannot see the gods face to face until we have faces.” Right now, some companies are able to offer an AI face to their users, but most do not. NLWeb is a chance for every company to have an AI interface (or simply “face”) for not just their human users but any bot that chooses to visit.

NLWeb is fully compatible with MCP and offers existing websites a simple mechanism to add AI search and other services to an existing web frontend. We put together our demo AI search frontend for O’Reilly in a few days. We’ll be rolling it out to the public soon.

Give it a try

Read More

We need a new…database?

Benn • Benn Stancil • May 16, 2025

Technology•Software•Databases•DataAnalytics•AI•Essays

On one hand, maybe calling Databricks a $36 billion mistake was a bit hyperbolic. Since I said that three years ago, the Nasdaq has crashed twice, and Snowflake, Databricks’ chief competitor, has gone from being a $70 billion company to a $60 billion company. Meanwhile, Databricks raised another funding round in January that valued the business at $62 billion. Uh, lol, whoops.

On the other hand, maybe it wasn't hyperbolic enough, and maybe the entire cloud database market was a mistake?

Like, if this blog is anything, it’s a stylized history of the last few years of data (and other concerns)—and here are some parts of that history:

Concern 1: Databases are really good now.

Companies collect tons of information about how their businesses work. They keep a ledger of customers’ purchases; they track clicks on their websites; they record your call for quality and training purposes. In the strange new world of the internet, we all emit billions of bits of structured digital exhaust—a like on a TikTok, an ad impression, a credit card swipe—and companies log it all.

Historically, this stuff was scattered across dozens of disparate systems. Today, it’s more centralized. Data is collected from a variety of sources, tidied up and cleaned, and carefully placed into a library of tables. If you were an analyst working for, say, the Charlotte Hornets, you could log into a single database and type, “Show me all the people we sent a marketing email to, and tell me if they bought a ticket to a game, and if they did, did we win the game, and did they buy any concessions?” Despite all of that data coming from different places, the magic of the entire modern data apparatus was that you could pretend that they didn’t.

Sure, this all a dramatic oversimplification, and nothing ever quite works this way in practice. Tables are rarely that well organized, there are often thousands of them, and they often overlap in confusing and contradictory ways. They’re frequently broken and out of date. And the questions people ask usually return messy answers: “Eh, wait, are these actually the people we sent marketing emails to? This doesn’t look right. Are we sure we’re logging this correctly? Oh, I think we used different campaigns for people who hadn’t been to a game before. No, that’s not it. Oh, no, that is it, we just did it wrong, and some people got both emails. But why does it say that this person who didn’t get an email clicked on one? Why does it say that this person who didn’t get an email clicked on 4,000? What is happening? Ah well, at least I don’t have to worry about whether or not we won the game.”

Nevertheless. That was the idea—a single pane of glass, for viewing all of your data—and over the last ten years, databases got a lot better at supporting it. They can hold, for all intents and purposes, unlimited amounts of information. They can run calculations over that data at nearly unfathomable speeds. They can be queried with many different languages. People can build custom apps on top of them. Or as I said a few years ago, comparing 2022 data stuff to 2012 data stuff:

The tools we have today—built and supported by thousands of people across dozens of companies—represent a profound leap forward from what we had then. And their effect extends beyond easing the daily frustrations of existing data scientists; they also made the work we did in 2012 accessible to a far greater range of companies and aspiring analysts and analytics engineers. Nearly every part of the industry is breathtakingly easier, faster, more powerful, and more reliable than it was a few short years ago.

….

Read More

Shared Memory, Knowledge Arbitrage, Why the Future of Creativity is Bright & AI Education FTW!

Implications • May 20, 2025

Technology•AI•Education•Innovation•Creativity•Essays

In this edition, we explore the evolving landscape of collective memory, the concept of knowledge arbitrage during platform shifts, the transformative impact of AI on education, the future of creativity, and share some off-the-record insights.

Me Becomes We: The Implications of Collective Memory

As AI-native companies develop new workflows, individual contributions are increasingly enriching the organization's collective memory. Tools like Glean, Notion, and Atlassian's Rovo facilitate enterprise-wide search across various documents and data, while platforms like Granola capture meeting notes. These tools are designed to integrate individual work into the organization's knowledge base, making it accessible to all members.

Consumer technologies such as ChatGPT and Claude are extending their memory capabilities, and it's anticipated that users will soon have the option to share their AI's memory with family and friends, similar to sharing a photo album or playlist. Looking ahead, as large language models (LLMs) remember our schedules, conversations, preferences, and purchases, sharing selective access to this memory with others could lead to a profound interconnectedness. This would enable loved ones to leverage—and even inherit—our accumulated knowledge, enriching their interactions and decisions.

In professional settings, colleagues could mine each other's interactions and insights, fostering unprecedented networking opportunities to advance business objectives. This evolution begins with shared memory through syncing data or selecting context windows (memories organized by theme) across families or teams. Ultimately, it could lead to literal brain melds when technologies like Neuralink enable direct brain-to-brain communication.

The Moment of Knowledge Arbitrage

We are at a pivotal moment of knowledge arbitrage during platform shifts. As new platforms emerge, early adopters can leverage their unique insights to gain a competitive edge. This period offers opportunities for those who can quickly adapt and harness the potential of these platforms to access and disseminate knowledge more effectively.

The Vastly Underestimated Impact of AI on Education

AI is transforming education in ways that are often underestimated. Beyond automating administrative tasks, AI is becoming a tool for personalized learning, adapting to individual student needs and learning styles. This personalization can lead to more effective teaching and learning experiences, fostering deeper engagement and understanding.

Why the Future of Creativity is Bright, Yet Different

The future of creativity is bright, albeit different from traditional notions. AI is not replacing human creativity but augmenting it, providing new tools and mediums for expression. This collaboration between human ingenuity and AI opens up new possibilities for innovation and artistic expression, leading to a renaissance of creativity.

Off-the-Record Insights

In addition to the above, we share some off-the-record insights that provide a glimpse into the future of technology and its implications for society. These insights offer a forward-looking perspective on the evolving relationship between humans and technology, highlighting the potential for growth, adaptation, and innovation in the face of rapid technological advancements.

Read More

Kara Swisher Hosts Fierce Debate On Future Of AI With Google VP, AI Ethicist

Youtube • Noema Magazine • May 15, 2025

Technology•AI•Ethics•Innovation•Regulation•Essays

In a recent discussion, Kara Swisher, a prominent tech journalist, engaged in a comprehensive conversation about the future of artificial intelligence (AI) with a Google Vice President and an AI ethicist. The dialogue delved into the rapid advancements in AI, its potential applications, and the ethical considerations that accompany its integration into various sectors.

The conversation began with an exploration of AI's transformative impact on industries such as healthcare, finance, and education. The Google VP highlighted AI's capacity to analyze vast datasets, leading to more accurate diagnoses in medicine, optimized financial strategies, and personalized learning experiences in education. However, Swisher raised concerns about the potential for job displacement due to automation, emphasizing the need for reskilling programs to prepare the workforce for new roles in an AI-driven economy.

Ethical considerations were a central theme in the discussion. The AI ethicist underscored the importance of transparency in AI algorithms to prevent biases that could perpetuate existing societal inequalities. They advocated for the establishment of ethical guidelines and regulatory frameworks to ensure responsible AI development and deployment. Swisher echoed these sentiments, stressing the necessity for global cooperation in creating standards that promote fairness and accountability in AI systems.

The conversation also touched upon the role of AI in content creation and media. Swisher expressed apprehension about AI-generated content potentially flooding information channels, making it challenging for consumers to discern credible sources. The Google VP acknowledged these challenges but highlighted ongoing efforts to develop AI tools that can assist in verifying information and combating misinformation.

In conclusion, the dialogue between Swisher, the Google VP, and the AI ethicist provided a nuanced perspective on AI's evolving role in society. While acknowledging the immense potential of AI to drive innovation and efficiency, they collectively emphasized the imperative to address ethical challenges and implement safeguards that ensure AI technologies are developed and utilized in ways that benefit humanity as a whole.

Read More

Sergey Brin, Google Co-Founder | All-In Live from Miami

Youtube • All-In Podcast • May 20, 2025

Technology•AI•Innovation•Leadership•Ethics•Essays

This interview with Sergey Brin, co-founder of Google, delves into the genesis, growth, and vision behind one of the world’s most influential technology companies. Brin shares insights on Google’s early days, emphasizing the foundational goal of organizing the world's information and making it universally accessible and useful. He reflects on the technological innovations that drove Google’s rapid expansion and how they have continuously sought to address complex information challenges through machine learning and AI advancements.

Brin discusses the evolution of Google's search engine and the importance of maintaining an extremely user-focused approach. He stresses that the company's success has hinged on relentless experimentation and learning from failures, fostering a culture that encourages innovation without fear. Notably, Brin highlights the impact of Google’s algorithms not only in search but in transforming broader digital landscapes like advertising, cloud computing, and hardware.

The conversation also explores Google's role in the broader tech ecosystem, including the ethical considerations involved in AI development and data privacy. Brin articulates a cautious optimism about AI's potential, emphasizing the need for responsible innovation frameworks to mitigate risks and ensure equitable benefits. He acknowledges societal challenges but is hopeful that technology, guided ethically, can contribute positively to solving global problems.

Brin touches on his current interests post-Google executive leadership, including projects at the intersection of technology and health, environmental sustainability, and longevity research. He envisions a future where tech continues to augment human capabilities and address critical challenges like climate change, healthcare, and education, leveraging AI and data science.

Key takeaways include:

The core mission of Google as an information organizer and the consistent user-centric innovation approach.
The pivotal role of AI and machine learning in evolving search and other digital services.
The culture of experimentation and learning from failure as central to Google's enduring creativity and growth.
Ethical vigilance in AI and data privacy as imperative for sustainable technological progress.
Brin’s contemporary focus on tech applications in health, environment, and longevity, reflecting a broader impact vision.

This conversation provides a nuanced understanding of Sergey Brin’s vision and leadership philosophy, illuminating how deep technical expertise combined with a pioneering mindset enabled Google’s transformational impact. It also underscores ongoing challenges and responsibilities the tech industry faces as it shapes the future.

Read More

How Kara Swisher Scaled Even Higher

Nytimes • Jessica Testa • May 20, 2025

Business•Management•MediaEntrepreneurship•Podcasting•MediaIndustry•Essays

Kara Swisher’s interviews made her famous among technology obsessives decades ago. She persuaded the rivals Bill Gates of Microsoft and Steve Jobs of Apple to play nice onstage. She reduced Meta’s founder Mark Zuckerberg, then just 26, to a puddle of sweat. She shoved her camera in the face of her future boss Jim Bankoff, Vox Media’s chief executive, among others.

But it wasn’t until she began podcasting that she reached an audience far beyond the tech world.

In 2018, she started “Pivot,” a news-chat podcast, with Scott Galloway, a serial entrepreneur and marketing professor who now has his own slate of brash business podcasts under the name “Prof G.”

They were an odd couple — she was grouchy, he was raunchy — but their banter was tender and intellectual when they weren’t torturing each other. Fans began stopping Ms. Swisher in public, recognizing the aviator sunglasses that had become a swaggering signature.

“I’d never made a product or a news thing that people thanked me for,” Ms. Swisher, 62, said in a recent interview at a cafe in the shadow of the National Cathedral in Washington, where she lives with her wife and children. “At the end of this long career, it’s like: ‘Oh, wow. I make something people really like.’”

So she and Mr. Galloway decided to assess its worth, shopping their portfolio of five podcasts around to other companies before their contract with Vox Media, their publisher, neared its end.

Competitive offers came in with guaranteed payments of about $40 million on four-year contracts, Ms. Swisher said. But in the end, they agreed to re-sign with Vox Media, with an unusual twist.

The deal does not carry any guarantees or upfront cash. The payday for Ms. Swisher and Mr. Galloway is instead based entirely on how much money their podcasts generate. Vox Media will pocket about 30 percent, while the co-hosts split the rest.

At the high end of back-of-the-envelope calculations — Mr. Galloway said the podcasts could generate $100 million in revenue over the four years — the pair would stand to make about $70 million excluding some costs. (A portion of the costs for their slate of shows is split among the hosts and Vox Media.)

The novel structure of the deal cements Ms. Swisher’s reputation for betting on herself. But it is also the kind of deal that could have wider implications, as more journalists follow Ms. Swisher’s example in fashioning themselves as new media entrepreneurs.

Ms. Swisher’s path to celebrity — a power broker who name-drops other power brokers — has taken her from The Washington Post to The Wall Street Journal to The New York Times, where she was an opinion columnist and host of a podcast called “Sway.”

Along the way, she co-founded two media businesses, AllThingsD and Recode; published three books; survived a mini-stroke; raised a family; and harbored few regrets. (Here’s one: “I was too nice to Elon for too long,” Ms. Swisher said of Elon Musk, the Tesla chief executive.)

She also learned a fundamental truth about herself: She does not want to be an employee, nor does she want to employ anyone. She wore a sweater to a White House Correspondents’ Association dinner party that warned people, or perhaps boasted, “I’m not for everyone.”

“Every day I get to decide what I do,” she said, “and it’s not dependent on anybody.”

….

Read More

The Agentic Web and Original Sin

Stratechery • Ben Thompson • May 20, 2025

Technology•AI•AgenticWeb•DigitalPayments•InternetEvolution•Essays

The concept of the "Agentic Web" represents a transformative shift in how artificial intelligence (AI) interacts with the internet. Unlike traditional AI applications that require human intervention, the Agentic Web envisions AI systems operating autonomously, making decisions, negotiating deals, and innovating without constant human oversight. This evolution is poised to reshape the digital landscape, enabling businesses to function more dynamically and efficiently. (forbes.com.au)

A critical component in realizing the Agentic Web is the integration of digital payments. The current reliance on advertising as the primary revenue model for online content has led to privacy concerns and a fragmented user experience. By incorporating seamless digital payment systems, a new content marketplace can emerge, allowing creators to monetize their work directly and users to access content without intrusive ads. This shift could address longstanding issues associated with the advertising-driven internet model. (theatlantic.com)

However, the transition to an Agentic Web is not without challenges. The current internet infrastructure is deeply intertwined with advertising, making a shift to digital payments complex. Additionally, concerns about data privacy and security must be addressed to build trust among users and content creators. Despite these hurdles, the potential benefits of an Agentic Web—such as enhanced efficiency, autonomy, and user experience—make it a compelling vision for the future of the internet.

Read More

The Anti iPhone

Spyglass • M.G. Siegler • May 22, 2025

Technology•Hardware•AI•VoiceComputing•ProductDesign•Essays

Yesterday, upon reading all the coverage about OpenAI buying io – which I'm going to continue to style as 'IO' for my own typing sanity – and especially watching the um, unique, video that Jony Ive and Sam Altman released to both formally announce IO and to talk a bit about what they're working on, it became pretty clear, pretty quickly that despite all the rumors to the contrary, the device wouldn't be a wearable.

Ive himself gave an interview just a couple weeks ago, hinting at what such a collaboration could be working towards – notably, a potential anecdote to smartphone addiction that Ive feels some level of remorse for having helped to usher into existence via his iPhone designs. Beyond that, all we get are whispers of "headphones" and "cameras" from WSJ sources. NYT talks vaguely about "ambient computing". What might this be? It's almost impossible to say, but I'm not sure it's exactly wearable, which is interesting. One thing it's not: a pair of smart glasses.

The key tell was the part of the video where Altman starts talking about how we're using the current "legacy" products, to use Ive's wording, to access future technology in the form of AI. It starts around 3:50 in:

"We have like magic intelligence in the cloud. If I wanted to ask ChatGPT something right now about something we talked about earlier, think about what would happen. I would like reach down, I would get out my laptop, I would open it up, I'd launch a web browser, I'd start typing and I'd have to like explain that thing and I'd hit 'enter' and I would wait and I would get a response. And that is at the limit of what the current tool of a laptop can do. But I think this technology deserves something much better."

Now, all of that could imply a wearable of some sort. But the key thing Altman seems focused on there is input, not necessarily the need to use it on-the-go. To me, this implies voice is the key of this device. Perhaps I'm biasing myself as I've been writing about this notion for years and years at this point. While the initial wave of devices leveraging Siri and Alexa got us closer to this type of computing, the truth is that the underlying tech wasn't good enough. Not nearly. We got sort of tricked because the voice recognition had finally gotten to the point where it worked well most of the time. But that actually wasn't the hard part, as it turns out. The hard part was the AI. OpenAI built the hard part first.

As I wrote almost exactly a year ago, right after the launch of GPT-4o and alongside it, the new voice mode for ChatGPT:

There were several points during OpenAI's demonstration of their new 'GPT-4o' model yesterday where I had to laugh. Not necessarily a "that's funny" laugh but more a "that's amazing" laugh. A profound laugh. A laugh to myself. A "this is it" laugh.

….

Read More

Venture Capital

The 10-Year Trend At Series B

Crunchbase • May 20, 2025

Business•VentureCapital•SeriesBFunding•ArtificialIntelligence•SiliconValley•Venture Capital

Over the past decade, Series B funding has experienced significant fluctuations, reflecting broader trends in the venture capital landscape. In 2014, the median Series B round was $11.7 million, with an average of $16.3 million. This figure saw a substantial increase, peaking in 2021 at a median of $32 million and an average of $46 million. However, by 2023, these numbers had declined to a median of $28 million and an average of $40 million, indicating a contraction in funding sizes. (xlera8.com)

The concentration of venture capital has also shifted notably. In 2024, the artificial intelligence (AI) sector dominated global venture funding, attracting approximately one-third of the total, amounting to over $100 billion—a significant increase from $55.6 billion in 2023. This surge underscores AI's growing prominence in the investment landscape. (vccafe.com)

Geographically, Silicon Valley maintained its dominance, securing $90 billion in venture capital in 2024, which accounted for 57% of global funding. This concentration is attributed to the region's strong AI presence, proximity to major tech companies, and a well-established startup ecosystem. (vccafe.com)

In terms of investment activity, the first half of 2024 saw an uptick in the number of companies raising funds and the number of deals done across Series A and B rounds. This trend suggests a renewed investor confidence and a more favorable environment for startups seeking growth capital. (ainad.net)

Overall, the past decade has been marked by significant volatility in Series B funding, influenced by sector-specific booms, regional dynamics, and broader economic factors. While the peak funding years have passed, the landscape continues to evolve, with AI and established tech hubs like Silicon Valley playing pivotal roles in shaping the future of venture capital investments.

Read More

The Production Capital Mosaic

Venturedesktop • Brett.Bivens • May 21, 2025

Business•Strategy•ProductionCapital•IndustrialInnovation•InvestmentModels•Venture Capital

A few weeks ago, Will Godfrey (CEO, Tangible) and I co-wrote a piece titled The Rise of Production Capital. For several years, I’ve used “Production Capital” to refer to a wide and growing range of approaches for financing the emerging physical technologies aimed at transforming critical industries like energy, aerospace and defense, manufacturing, materials, and transportation.

The term is a nod to Carlota Perez’s seminal framework for understanding technological revolutions, particularly the constructive deployment “golden ages” that follow from periods of i) technological “installation” and ii) intense socio-political tension and conflict, wherein:

Constellations of convergent innovations create the building blocks that enable a new class of innovators to transform key industries.
Technology is rapidly adopted across industries, extending its reach beyond geographically limited innovation hubs to benefit a wider segment of society.
Capital flows into real businesses contributing fundamental value to the economy.

This magnitude of the deployment age is captured by BlackRock’s estimate of the impending $68 trillion global infrastructure investment boom, unfolding in an era increasingly defined by what Russell Napier calls “National Capitalism”, where state priorities reshape capital allocation toward energy security, industrial capacity, and technological sovereignty.

With that (still broad) framing in place, I thought it would be helpful to sketch out a mosaic of the types of firms and companies I’ve seen focused on this. The use of the term “mosaic” is intentional. This isn’t a comprehensive market map or a clean framing of the so-called “capital stack”.

The boundaries between these models are often blurry. Most are still a work in progress, not yet fully legible to the institutional capital allocation world that cements the clean lines between asset classes and financial products over the long term. These organizations are also, to lean on the scientific definition of mosaic, typically “composed of cells of two genetically different types” – founder x financier; venture x infra; physical x digital.

This entrepreneurial nature of Production Capitalists is captured well by Perez:

"Financial capital can successfully invest in a firm without much knowledge of what it does or how it does it. Its main question is potential profitability, sometimes even just the perception others may have about it. For production capital, knowledge about product, process, and markets is the very foundation of potential success."

Carlota Perez, Technological Revolutions and Financial Capital

That these organizations transcend straightforward categorization is what creates their opportunity.

This is also an exercise to understand what I am missing. So if I have omitted categories or interesting approaches, please let me know!

Things Shaped Like an Investment Firm

Venture + Capital Markets Connectivity → "Unlock downstream financing" has always been a core VC job to be done. Diverse downstream capital needs arrive early in the physical technology company life cycle, creating an opportunity for specialist firms to earn a spot on the cap tables of the best emerging industrial companies by building a capital markets function on behalf of their portfolio. Some venture firms do this well informally, but few (none?) have developed this as a systematic platform capability. By helping founders optimize around structure, strategy, and capital markets connectivity, firms can have a substantial impact on the metrics that matter to them, like dilution over time (equity efficiency) and set companies up for more efficient subsequent financings.
“GAP” (Growth and Project) Capital → The "Compound Startup" of the Production Capital universe. Vertical integrators, building a system of capabilities that spans venture, project development, financial engineering, and industrial business development to power companies through the proverbial valley of death. The problem is abundantly obvious, but the ability to commandeer and coordinate the capital and cultural resources (i.e. world world-class talent across several disciplines from day one) to make this model a reality – as The New Industrial Corporation has done – is hyper-scarce. (h/t Will Dufton from Giant Ventures for the term GAP capital)
Modern Merchant Banks → A more flexible, “fundless” variation of the first two models, leaning into the diverse nature of capital problems to be solved inside new industrial businesses. Bespoke capital support via advisory and strategic positioning, systematic capital markets connectivity, and investment (via balance sheet or SPVs) – mirroring the way entrepreneurial financiers of previous eras built centrality during industrial shifts.
SPACs and PIPEs → Will value distribution in the new industrial ecosystem more closely resemble Silicon Valley (power law) or the German Mittelstand? If it is the latter, and if the barriers to going public remain for mid-sized companies (roughly $100m - $10b in value) we might see a revival of SPACs – ideally wielded more responsibly – to raise large amounts of capital needed for industrial scale-up, align with strategic investors, simplify access to government programs, and use public equity as M&A currency. Nuscale is a rare success story here, while blank check companies like Perimeter are emerging with this angle in mind.
Venture Turnarounds → Industrial technology scale-up and venture equity are not always perfectly compatible (as this list conveys). As more venture capital has flowed into the physical economy, more companies are finding themselves in a position with strong IP, physical assets, and commercial traction but misaligned cap tables, organizational structure, and operating models. This creates an opportunity to restructure and reaccelerate these businesses more sustainably. Jeremy Giffon discussed this on TBPN, and I am aware of a few efforts behind the scenes, but haven’t seen any public announcements.

Read More

New SaaStr on 20VC! “The VC Playbook: What’s Working in 2025”

Saastr • Jason Lemkin • May 16, 2025

Technology•AI•VentureCapital•Fundraising•Leadership•Venture Capital

New SaaStr on 20VC! “The VC Playbook: What’s Working in 2025”

We’ve back this week on 20VC with Harry and Rory from Scale for another great deep dive on the latest in VC funding and more. We kick off with a quick deep dive on Vertical SaaS leader for restaurants Owner raising at $1 Billion — a deal SaaStr Fund led at seed. Why was it so hot?

5 Key Takeaways

Capital is plentiful for the right metrics: Companies growing 10%+ MoM with $40M+ ARR can still close $100M+ rounds in days, while Series A rounds overall are down 81%. The divergence between haves and have-nots has never been greater.

AI is creating urgency but also fear: CMOs and other executives are buying AI tools out of fear of becoming obsolete, creating a gold rush for companies like Clay that position themselves as job-saving solutions.

The trillion-dollar investment thesis: VCs are now evaluating deals based on “odds of trillion-dollar outcome” rather than traditional metrics, justifying seemingly irrational valuations (like Perplexity at $14B) for companies with even a small chance at massive outcomes.

Non-technical CEOs can win in AI: Despite conventional wisdom, OpenAI is dominating with non-technical leadership because they excel at recruitment, empowerment, and strategic partnerships.

Winning today requires aggressive capitalization: The playbook for category winners is to raise more than needed and “scorched earth” the competition, as the pace of innovation and competition in AI-enabled categories is unprecedented.

Why Series A Is Hard (But Great Companies Still Get Funded)

Reports show Series A rounds are down 81% – but this is actually normal. The seed round is the “believe in the team” round, while Series A is the “show me the traction” round. And traction is hard to manufacture.

The key insight: Don’t go fundraising for a Series A unless you’re certain you have what it takes. Smart companies sometimes choose to delay raising until they have undeniable metrics.

For founders: If you’ve built a B-tier startup that could get funded in 2021 but can’t today, that’s just market reality. The bar has gone up. The best startups with stellar metrics still get 5+ term sheets.

The New Fundraising Timeline: Preempts Getting More Aggressive

The competition for quality deals has become so intense that Series A investors are asking “how quickly after the seed can we preempt?” Often this happens just 2-3 months after a company’s seed round.

Why this works: If a company has grown 50% in just two months since you first met them and you’ve already done your diligence, why wouldn’t you move? The best companies in high-growth markets have the highest velocity of fundraising.

Hard truth: If you’re not tracking your top 10-20 companies you want to invest in over the next 12 months, you’re doing it wrong. The days of “wandering around hoping stuff will turn up on Monday that will make you money by Wednesday” are over.

….

Read More

Why VC and software have PE envy

Mtb • Matt Brown • May 13, 2025

Business•Startups•VentureCapital•PrivateEquity•VerticalSoftware•Venture Capital

Imagine you’re the world’s most entrepreneurial dentist, the Mark Zuckerberg of molars. You’re great at cleaning teeth, but you’re also great at running a practice. You have your own playbook for the dental business.

Thanks to your playbook, your practice is larger and more profitable than the market average. You’ve surveyed your dentist friends. You’ve learned that you’re the best at acquiring patients, running your back office, configuring your CRM to minimize cancellations, and getting paid quickly. But you aren’t satisfied with cleaning teeth all day. You want to build a dental empire. The gap between the average business and what’s possible with your playbook is your opportunity.

How do you start your empire? Until recently, there were two broad and distinct paths: private equity (PE) and venture capital (VC).

In the PE model, you’d raise money to acquire a controlling stake in one or more practices. Then you would have full control to implement your playbook and manage those practices day-to-day.

In the VC model for B2B software, you’d raise money to start a company that builds software that codifies your playbook for any practice to implement. Maybe it’s a suite of CRM, scheduling, and billing tools, with opinionated workflows and design choices specifically for dental practices. You wouldn’t own or manage any practices directly, but your software would help them run better.

Both models start with the same principles: (1) the average company in a given vertical isn’t run as effectively as it could be, and (2) you have the insights and playbooks to run the average company more effectively. You believe you can close the gap between the blue and green dots, and get paid handsomely for it.

However, the PE and VC models take very different approaches to point 2. The tradeoffs of each model are most obvious in their approaches to control, concentration, and value capture.

The PE model is high control, high concentration, and high value capture. It says, “This specific business can be run better, and so it’s likely undervalued. I’m going to buy and run it, applying my playbook. As the business grows and becomes more efficient, it will become more valuable. I’ll benefit from the increased equity value.”

The VC model (at least within B2B software) is low control, low concentration, low value capture relative to an individual business. It says “There are lots of businesses in this market, and most of them could be run better. I’m going to build software that helps them do that. I’ll charge a small fee, but will sell it to many businesses.”

The PE and VC models are extreme ends of a spectrum, rather than distinct models. PE and VC firms are both in the business of generating returns for their investors. They take capital, combine it with their belief in the superiority of their playbook, and then implement their strategies and playbooks in a given vertical. This may involve buying businesses or building tools to enhance their performance, generating profits, and returning the profits to their investors.

Until recently, you’d be forgiven for thinking PE and VC are effectively distinct entities. However, headlines are making it more apparent that they’re just a spectrum of strategies and that there’s a lot of white space in between them.

Why are VCs (and the companies they fund) adopting PE-like strategies? Let’s explore.

In seeking outsized returns, VCs and venture-funded startups are venturing beyond their previously narrowly defined model. They’re getting more creative and aggressive in exploring the messy middle, the whitespace between the previously distinct PE and VC ends of the spectrum.

Several factors are behind this move towards the messy middle. Some are pushes from the traditional VC model: the classic, almost boutique approach of minority investments in asset-light, high-growth, all-or-nothing, power-law-seeking, software-first-and-only startups. As this model gets more competitive, smart founders and investors realize they must try something different. At the same time, several powerful factors are pulling toward these new models: new technologies and business models seem to enable venture-like returns from traditionally non-venture-type businesses.

The push side is well-documented: the traditional VC asset class has become saturated, especially in B2B SaaS. Nearly a trillion dollars have flowed into the venture industry in the last decade. This has led to a proliferation of companies serving every vertical and every niche.

At the same time, AI promises to further reduce the cost of software development. That isn’t to say that the B2B SaaS market is going to zero, or that there won’t be another generational B2B SaaS business. But the noise and saturation make it harder for these companies to grow quickly while retaining customers and high margins. There aren’t many land grab opportunities in pure SaaS like there were in the 2010s.

Even for companies with great products, it’s getting harder to sell to businesses with SaaS fatigue. To return to the original dental example: suppose you (the entrepreneurial dentist) build the best vertical software for dental practices. Your target customers are already inundated with such pitches. The same is true if you’re pitching the most well-oiled acquisition strategy for underperforming practices. It’s hard to sell them a genuinely better product, and it’s also hard to convince them to sell their businesses.

Read More

If 198 Pieces of Unsolicited, (Possibly) Ungoogleable Advice for Founders Were Not Enough

Cupofzhou • David • May 17, 2025

Business•Startups•Fundraising•Governance•Hiring•ProductManagement•Venture Capital

This is my third iteration of the 99 series for founders. You can find the first two here and here. The premise for this series was simple. The best, most insightful, unsuspecting lessons are hidden in the deepest, darkest corners of the internet. Hell, many more are hidden in rooms behind closed doors. The goal of this 99 series is to unveil those. Advice you’ve likely never thought about, and most likely have never heard of.

While you don’t need to read all the below at once, it’s helpful to keep the below at your fingertips for when you do need them. As always, unless the advice is not cited, all advice has been backlinked to its source, in case you want the longer, sometimes more nuanced version.

To make it easier for you, I’ve also pooled the advice in categories, depending on your needs:

Fundraising
Governance
Hiring/Team/Culture
Product/Customers
Competition
Legal
Expenses
Secondaries

P.S. Have I started the next one in the 99 series for founders? Yes, I have. Stay tuned!

Fundraising

1/ “Once you take venture capital, the venture capitalist’s business model is your business model. You’ve got to get liquid at a number that makes sense for them. High valuations are good because you take less dilution. Et Cetera. But the reality is that when you have a high valuation, that starts to eliminate your options.” — Chris Douvos

2/ The employee option pool is easier to negotiate than asking an investor to take less ownership. The pool at the time of term sheet comes out of founder/team’s equity. If the pool becomes completely allocated post-investment, you need to go back to the board and ask for a larger pool, and everyone (you and VCs) gets diluted then.

3/ Beware of the “senior pari-passu,” which means that that investor gets paid back before everyone else on the preference stack AND they get equal footing with all the other investors. If one investor has this, subsequent rounds may demand the same.

4/ Repeat founders often ask for co-sale right immunity (usually 15%) when putting together term sheets. Co-sale rights give investors first dibs on buying your equity if you sell before a liquidity event, and they can sell alongside you. This can send negative market signals.

5/ If any corporates own more than 19.5% of a company, they must treat your company as a subsidiary for accounting, making them less valuation sensitive.

6/ You’re likely not the only one in market with your solution. A competitor raising a massive round signals market validation, not necessarily a reason to change your pitch. Only change your pitch if customers prefer the competitor.

7/ “Once you have $500k+ raised, spend 2/3 of your time on funds, 1/3 on small checks.” — Ash Rust

8/ Beware of SAFE overhangs. Avoid raising more than 25% on SAFEs compared to the next priced round. — Martin Tobias

9/ Don’t say “The market is so large, there are room for many winners.” To a VC, that sounds like you’re being beaten by competition. — Harry Stebbings

10/ If many employees lack startup experience, carefully decide how and when to be transparent about company realities; startup experience level will influence whether transparency helps or hurts. — Javier Soltero

11/ To fundraise, even if your recent months were tough, show three months of strong growth before fundraising. — Jason Lemkin

12/ Revenue growth benchmarks for VCs: before $1M ARR, grow 10-15% monthly; around $1M ARR, 8-10%; around $10M ARR, ideally doubling.

13/ “An investor is an employee you can’t fire.” — Vinod Khosla

14/ “Things that break the rules have a bigger threshold to overcome to grab attention but tend to have stronger, more dedicated followings. Blandness gets fewer dedicated followers.” — Brandon Sanderson

15/ Great worldbuilding with poor characters/plot is just an encyclopedia; great characters and plot with weak worldbuilding can still be excellent. This applies to markets and teams as well.

16/ In all great stories, the protagonist (pitch) is proactive, capable, and relatable. Your pitch should show at least two of these. — Brandon Sanderson

17/ “Data rooms are where fundraising processes go to die.” Prioritize live conversations, ask for 15 minutes on investor calendars to prepare info. If they refuse, you’ve lost the deal. — Mark Suster

18/ “Second conversation with a serious investor explores what you’re trying to prove and to whom.” — Fund III GP

19/ “Set your own agenda or someone else will.” — Melinda Gates

20/ “The ‘raise very little’ strategy only works if the market is believed tiny or unimportant. If others pay attention, you must beat the next guy.” — Parker Conrad

21/ Beware of stacking SAFEs, model founder ownership to stay above 50% pre-Series A. — Itamar Novick

….

Read More

The Bar Today for a Series B

Saastr • Jason Lemkin • May 17, 2025

Business•Startups•Funding•SaaS•SeriesB•Venture Capital

In the world of SaaS startups, securing Series A funding is an achievement worth celebrating. At least for a day :). It’s a sign you’ve hit Initial Traction.

But each round usually gets harder. Successfully graduating to Series B takes a lot of strong growth and metrics.

The latest data from Carta analyzing 10,755 US Series A startups offers a revealing look at these graduation rates – and there are important lessons for B2B founders to absorb.

The comprehensive Carta data reveals several critical insights about Series A to B progression:

Time Horizon Matters Tremendously: The probability of securing Series B funding increases substantially over time. For cohorts with sufficient maturity (2018-2020), we see progression rates reaching 40-50%+ by Year 4, compared to just 1-4% in the first quarter after Series A.
Vintage Effect Is Real: Startups that raised Series A between 2018-2020 show higher Series B graduation rates than those raising in 2021 or later. The 2020 Q1 cohort shows particularly strong performance, with over 55% reaching Series B by Year 4.
The Recent Funding Winter: There’s a visible cooling in graduation rates for startups that raised Series A from 2021 onwards, with significantly lower progression percentages across all time horizons.
Signs of Recovery: As the chart annotation indicates, there are “signs of life” with graduation rates “inching back up in more recent cohorts” – a cautiously optimistic indicator for the current fundraising environment.

What This Means for SaaS Founders

Plan for the Long Haul
The data makes it clear: the path to Series B is a marathon, not a sprint. For SaaS companies specifically:
Most successful Series A companies take 2-3 years to secure Series B funding
The median time between rounds is approximately 24 months
Only a small minority (typically under 10%) secure follow-on funding within the first 6 months

Action item: Structure your Series A round to provide at least 24-30 months of runway, with contingency plans for extending it if needed.

Understand the New Metrics Bar for SaaS Series B. The funding environment since 2021 has reset expectations. Today’s Series B SaaS companies typically need to demonstrate:
$4-8M in ARR (up from $2-4M pre-2021)
2-3x YoY growth (with some flexibility for higher efficiency)
Clear path to profitability with improving unit economics
CAC payback periods under 18 months
Net revenue retention above 110%

Action item: Build your post-Series A strategy around achieving these specific benchmarks rather than vanity metrics.

Capital Efficiency Is the New Growth
The data shows that 2021 was a turning point, with graduation rates dropping significantly. This coincides with the shift from “growth at all costs” to “efficient growth”:
Companies that raised Series A in 2020 or earlier could often secure Series B primarily on growth metrics
Post-2021 companies face stricter scrutiny on burn rate and capital efficiency….

Read More

FOUNDERS: stop reimbursing your VC’s legal fees

Auren • Auren Hoffman • May 22, 2025

Business•Startups•VentureCapital•LegalFees•FounderFriendly•Venture Capital

FOUNDERS: stop reimbursing your VC’s legal fees

I remember the terrible feeling after wiring $38,432 to a law firm that did not represent me ... it represented the venture capitalists negotiating against me. Everything about it felt wrong.

Ramen-eating founders should not be footing legal bills for billionaire venture capitalists. Yet that’s still the industry default.

The most outrageous term-sheet item is also the most common. Almost every VC asks their founders to reimburse their legal fees. I’ve raised more than 15 priced rounds in my career and paid my VC’s legal fees every time.

Any VC that does this is the opposite of founder-friendly. VCs should pay their own legal fees.

And these legal fees aren’t cheap – it is not uncommon for a venture capitalist to ask for a $50k to $100k reimbursement on a $5 million round. That’s like hiring a good software engineer for six months—gone to lawyers in a single wire. It’s a charge to the companies of 1-2% of the round right off the bat.

These fees are very dilutive to the company and they actively incentivize lawyers on both sides to rack up billable hours. Most of these deals have a legal fee cap but guess what happens when you set a cap? The lawyers work hard to meet it.

VCs: “We back scrappy underdogs.”

Also VCs: “Pay our legal fees which costs more than your salary.”

Parkinson’s Law says that work expands so as to fill the time available for its completion. You will never, ever get a bill from the opposing counsel’s lawyers that is less than 90% of the legal fee cap. The VC’s lawyers have no one watching their costs. And every $2,000 they bill is usually at least a day delay in closing.

Founders should not be funding VC legal fees.

It's like inviting a poor friend over for dinner and charging them for the ingredients.

Double-Taxing Shareholders and LPs

This practice is not just unfriendly to founders … it is also unfriendly to LPs.

“Our endowment shelled out over $3 million last year for ‘deal expenses’ we never approved,” confessed one large university LP.

By having portfolio companies pay for a VC’s legal expense, the VC is essentially backdooring another expense that LPs will never see or audit.

This is because VCs are billing the legal fees to the fund without explicitly telling their LPs about these billing practices. That is just wrong.

Most people don’t realize that venture firms split their expenses between the management company, which collects the ~2% annual fee, and direct charges to LPs in their funds. Costs like deal diligence, fund administration, banking, annual meetings, and even travel are sometimes passed through to LPs. Meanwhile, the 2% management fee is carefully protected to fund large GP salaries.

So why not just pass the legal fees to the LP? Well, VCs are required to report fees passed through to LPs, and these fees face scrutiny (because they are additional fees that come directly out of the LP’s pockets) and must go through a deep audit.

So the VCs elect to backdoor the fees.

It is kind of unseemly.

And it might not even be legal.

Most VCs already over-bill their fund for things they shouldn’t and overly hurt their LPs. But at least most of these expenses are explicit and categorized and the LPs can see them. Hidden fees to the fund are against the spirit of being a limited partner.

Expensing a VC’s legal fees takes advantage of BOTH the company’s shareholders and the VC’s shareholders. It is doubly bad.

It is both founder unfriendly and LP unfriendly.

Why This Bad Practice Persists

The only logical answer is because they can. Because no one fights back.

While 98%+ of VCs do this, not EVERY VC does this. When I did my first VC deals, I reflexively asked founders to reimburse my legal fees … but quickly ended the practice because it is extremely unfriendly to both founders and LPs. Plural (a highly regarded European VC), Flex Capital (where I work), Relentless, and Y Combinator all have an explicit policy of not asking their portfolio companies to cover their legal fees. If you know of other VCs that are truly founder-friendly, please comment or email me and I will add them here. We can celebrate them.

Founders—everything is negotiable. Start with these asks:

“Line-item ‘Legal Fees: $0’—the VC pays their own.”
Go further: add a “delay-penalty clause": for every extra week caused by the VC’s counsel, the VC reimburses the company $5k.

Who Else Can Fix It?

YC already popularized the SAFE which usually does not need legal fees because of its simplicity. YC should go one step further and specifically tell VCs that it will discourage its founders from working with investors that push founder-unfriendly terms in priced rounds.

Big LPs also can put pressure on VCs to change the practice. LPs should require VCs to report any backdoor fees to the funds.

Let’s change this bad practice.

Founders should not be funding VC legal fees.

It's like inviting a poor friend over for dinner and charging them for the ingredients.

Every VC’s slogan:

📝 “We’re founder friendly.”

Yet every VC term sheet:

📝 “Please reimburse our $100K legal bill.”

Cool cool cool.

any VC that tries to force companies to pay their legal fees, by definition, founder unfriendly. next time you get a term sheet like that, ask the VC if they are having money problems.

forcing little companies to pay legal fees is an insult.

VCs: “We back scrappy underdogs.”

Also VCs: “Pay our legal fees which costs more than your salary.”

HT: Lily Petherick, Rajal Patel, Sush Bhardwaj of Flex Capital.

Read More

Seed Investing At Scale | David Tisch, Managing Partner at BoxGroup

Youtube • Uncapped with Jack Altman • May 22, 2025

Business•VentureCapital•SeedInvesting•StartupFunding•Entrepreneurship•Venture Capital

David Tisch, Managing Partner at BoxGroup, discusses the evolution of seed investing and the strategies that have contributed to his firm's success. BoxGroup, established in 2007, has invested in over 500 seed-stage startups, including notable companies like Plaid, Ro, Ramp, and Stripe. (thetwentyminutevc.com)

Tisch emphasizes the importance of being an early believer in a startup's potential. He notes that securing the first investment is often the most challenging, yet it provides founders with the confidence and momentum to move forward. BoxGroup prides itself on being that initial supporter, aiming to be the second or third largest check in a round, with investments typically ranging from $250,000 to $1 million. (greenbayhotelstoday.com)

The firm's investment approach is collaborative, focusing on supporting founders without overemphasizing ownership stakes or board seats. Tisch highlights that BoxGroup makes about 40 investments annually, seeking to be part of founders' journeys without dominating them. (deciphr.ai)

In terms of portfolio construction, Tisch acknowledges the challenges of increasing ownership stakes in later funding rounds due to competition from larger firms. He suggests that maintaining ownership is more about building strong relationships with both founders and other investors rather than aggressively seeking higher stakes. (deciphr.ai)

Tisch also discusses the impact of multi-stage funds on seed rounds, noting that while larger funds can lead to increased competition, they also validate the market and attract more attention to the sector. He believes that the key to success in seed investing lies in identifying exceptional founders and supporting them through the various stages of their company's growth. (thetwentyminutevc.com)

Overall, David Tisch's insights provide a comprehensive understanding of the dynamics of seed investing, emphasizing the importance of early support, collaborative partnerships, and strategic portfolio management.

Read More

AI

AI Adoption Among Private SaaS Companies and Its Impacts on Spending and Profitability

Saas Capital • Nick Perry • May 16, 2025

Technology•AI•SaaS•Adoption•Profitability

AI Adoption Among Private SaaS Companies and Its Impacts on Spending and Profitability

Since the launch of ChatGPT in late 2022, AI has dominated both discourse and funding dollars in the SaaS industry. With opinions ranging from existential threat to limitless opportunity, we were keen to extend our annual survey to cover several questions on AI in order to glean just how it is being incorporated by private SaaS companies and what the impact has been to date on spend. Below are our key takeaways:

AI utilization is not limited to several large companies or young start-ups. Roughly three-quarters of our respondents are using AI in their day-to-day operations or product.

The majority of respondents are using limited AI functionality to complement their core software offering, but the role of AI is expected to expand going forward.

At the margin, AI utilization strategies to date appear to be supporting companies’ drive to profitability rather than a renewed emphasis on growth at all costs.

Teasing out the impact of AI on particular spending categories is not possible without an understanding of how each company is approaching its AI strategy and accounting for its AI spend.

Nevertheless, we do note the following broad conclusions when comparing companies with $1–20 million ARR that are using AI in operations and product vs. similarly sized companies using no AI. The former group of companies reported higher COGS as well as selling and marketing costs, and lower R&D and G&A expenses relative to companies not using AI.

AI Adoption has been Broad-Based and is Likely to Accelerate

As part of our annual survey, we asked respondents several questions about their use of AI both in their operations and in their product. Their answers reflect the breadth of AI adoption among our universe of private SaaS companies.

Over 76% of respondents indicated they were using at least some amount of AI in their existing products, and 69% are deploying AI solutions in their day-to-day operations.

Now, it bears stating that what is termed AI can mean vastly different things to different operators; nevertheless, the conclusion is unmistakable: AI (broadly defined) has been widely incorporated by SaaS companies.

This conclusion holds regardless of funding type (i.e., if the company is bootstrapped or equity-backed). There is a slight preference for bootstrapped companies to deploy AI in operations (70% of respondents) compared to equity-backed companies (66%), whereas equity-backed companies were somewhat more likely to have built AI functionality in their product: 79% vs. 71% for bootstrapped companies.

Interestingly, AI adoption in product appears to be the typical first step for companies. While we did not ask respondents to specify the timeline of their AI adoption in operations vs. product, we can derive an informed guess by comparing the number of companies using AI in their product that are also using it in operations and vice versa. Of the companies that deployed AI in their product, 50% were also using AI in their daily operations.

Conversely, 88% of companies that had AI in their operations were also using it in their product. This supports the conclusion that, in general, companies are first deploying AI in their product before implementing it in their daily operations.

If the breadth of AI adoption is not in doubt, its depth is a more nuanced matter. We asked companies to specify whether AI was a limited, significant, or predominant piece of their product. The majority of companies reported having limited AI functionality around their core software product. Interestingly, the same percentage of companies reported using no AI in their product as the combined percentage of companies that reported their product was either significantly or predominantly/entirely AI.

Again, controlling for whether the companies were bootstrapped or equity-backed does not meaningfully alter the results. Among bootstrapped companies, 52% reported their product used limited AI compared to 58% of equity-backed respondents. There is also a slight increase in the percentage of companies whose product is either significantly or predominantly/entirely AI (21% combined) among equity-backed respondents compared to 18% for bootstrapped companies.

Overall, it should not be surprising that companies have implemented a “limited” approach to AI adoption in their product so far. There are myriad reasons why AI utilization in product is limited so far, but the simplest explanation is that this takes time, and we are not far removed from AI bursting into the mainstream. Retooling solutions can include significant commercial, strategic, and engineering changes. There may also be lingering questions around fit, accuracy, utility, and security/privacy. All of these take time and energy to address. A snapshot of the data may look like companies prefer a limited approach to AI, but at least as of now, this is more likely a pass-through stage of an industry learning to walk with AI before it can run.

In fact, nearly 92% of respondents indicated that they planned to increase their use of AI in 2025.

With this in mind, let’s take a look at how the use of AI has impacted profitability and spending.

AI Adoption and Profitability

AI has the potential to unlock new approaches and efficiencies in how SaaS companies build, execute, and deliver their product, as well as how they go to market.

Effective AI adoption offers the opportunity to create more output for a given input. With this capability, SaaS operators have a choice to deliver the same output with fewer resources or maintain (and even grow resources) to deliver greater output.

Whichever path they pursue will have different results in terms of spend and profitability. Companies may use AI to drive toward profitability while maintaining output, or they may double down on growth. They may also choose to implement different strategies for different components of their business. For example, a company may utilize AI in its Research and Development to reduce resource requirements but simultaneously decide to bolster its Sales and Marketing teams with an eye to growing its market footprint. There is also the added complication of companies potentially allocating AI spend differently across operations or product.

Without knowing the individual priorities and accounting policies of each respondent, it is impossible to draw specific conclusions about the absolute impact of AI on profitability and spending. Nevertheless, we can compare the expense structures and operating profitability for AI adopters vs. non-adopters as a whole to see if any broad observations emerge.

Looking at AI adoption in SaaS product, there is no distinction in profitability between companies using AI and those that are not. Nearly 58% of companies using AI in their product reported operating at or above breakeven. For companies without AI in their product, the equivalent number is 59%. When it comes to usage in product, AI does not appear to be a determinant in aggregate-level profitability at this stage.

Where a difference in operating results does show up is when we look at whether SaaS companies are implementing AI in their day-to-day operations. Companies using AI in their operations were more likely to be operating at breakeven or profitability (61% of reporting companies) compared to companies that had not incorporated AI in their operations (54%). At the margin, this would support the idea that, at least so far, efficiencies generated by AI utilization in day-to-day operations have more than offset any increased spend related to AI adoption.

Controlling for whether a company is equity-backed or bootstrapped provides additional detail as to how AI may be impacting profitability. The first thing to note for this analysis is that bootstrapped companies are much more likely to operate at breakeven or profitably than their equity-backed peers.

Across the universe of survey participants, 82% of bootstrapped companies are breakeven or profitable compared to 43% of equity-backed companies.

For bootstrapped companies, implementing AI in product or daily operations does not materially alter the percent of companies operating at breakeven or above. Where the impact of AI may be seen more starkly is in the universe of equity-backed companies using AI in their operations or product. Forty-nine percent of equity-backed companies using AI in their operations were operating at breakeven or above (compared to the 43% baseline for all equity-backed companies regardless of AI use). More notable still is that for equity-backed companies not using AI in daily operations, only 29% reported being breakeven or profitable. A similar, though less pronounced, finding occurs when we examine equity-backed companies not using AI in their product. For this cohort, 39% reported operating at breakeven or better compared to the 43% baseline.

A plausible explanation for AI cost efficiencies seemingly manifesting in equity-backed companies more than bootstrapped companies is that, as there are a greater number of equity-backed companies operating at a loss, the realization of cost efficiencies through AI adoption in operations is more likely to push a handful of companies out of operating losses and into breakeven or profitable results. Since there are fewer bootstrapped companies operating at a loss, there are fewer opportunities for these companies to switch from operating losses to profits. A similar pattern may be playing out for AI utilization in product, although to a less pronounced degree.

Read More

The AI Warnings Will Continue Until Morale Improves

Nymag • John Herrman • May 17, 2025

Technology•AI•Workplace•JobDisplacement•CorporateStrategy

The AI Warnings Will Continue Until Morale Improves

Tobi Lutke, CEO of Shopify, recently sent a memo to his staff declaring that AI usage is now a “fundamental expectation” for everyone at the company and that there will be “usage questions” added to peer and performance reviews. He calls out “amazing colleagues” less to emphasize a sense of workplace camaraderie than to call attention to superhuman colleagues, “the kind who contribute 10X of what was previously thought possible” and who can now, by using AI tools that have “become 10X themselves,” get “100X the work done.”

If that math sounds like it might raise some questions about Shopify’s overall staffing levels, he got around to that, too. “Before asking for more Headcount and resources, teams must demonstrate why they cannot get what they want done using AI,” he writes. Lutke’s memo is stark and assertive. It portrays a boss becoming more comfortable telling his employees How Things Are and What Must Be Done, at least right up until the part about what happens after everyone makes the effort to “figure out” artificial intelligence: “AI will totally change Shopify, our work, and the rest of our lives. We’re all in on this! I couldn’t think of a better place to be part of this truly unprecedented change than being here.” Cheers!

The memo got instant coverage for its line about head count and went LinkedIn viral for its AI mandate. It also inspired some copycats, among them a dire memo from the CEO of gig-work platform Fiverr, Micha Kaufman, who preemptively shared his warning to employees on X “before it gets out somewhere else.” As a business, Fiverr is extremely exposed to AI tools as they already exist; its platform is full of freelancers offering low-cost translation, illustration, and basic software-development help.

This memo, however, wasn’t about that. It was addressed to and about Fiverr’s corporate staff. “I’ve always believed in radical candor,” Kaufman wrote. “AI is coming for your jobs. Heck, it’s coming for my job too. This is a wake-up call.” Where Lutke’s memo merely suggested belt-tightening and heightened expectations, Kaufman’s was more explicit. “Study, research, and master the latest AI solutions in your field,” he wrote. “You are expected and needed to do more, faster, and more efficiently now.” If you don’t use AI, he continued, “your value will decrease before you know what hit you,” and employees need to stop “waiting for the world or your place of work to hand you opportunities to learn and grow.” If you don’t like what he wrote, he said, “I honestly don’t think that a promising professional future awaits you.” Is everyone doomed? “Not all of us, but those who will not wake up and understand the new reality fast, are, unfortunately, doomed.” For now, it’s time to stop “disregard[ing] reality” and to get on the “winning side of history.”

These memos are less notable for what they say than for how they say it, and, perhaps most of all, for what they don’t contain: news about massive restructuring, layoffs, or fundamental changes to business plans.

Granted, announcements along those lines might indeed be coming: There is genuine widespread uncertainty and concern about white-collar AI job displacement, and a wide range of job types could be altered by generative AI. But until they do, these memos are serving more immediate goals. In a small way, they function as ways to let investors know that leadership isn’t behind the curve and won’t miss out on the next big thing. Mostly, though, they’re a way to signal to workers, who are already anxious about AI, that they should be scared about their jobs — and that, in the meantime, they should know their place. (Of course, we in the media are experimenting with AI mandates too.)

AI anxiety is epidemic in the economy, driven both by startling advances in tools that hundreds of millions of people can try for themselves and predictions by industry figures that massive disruptions are imminent. But perhaps nowhere is it as acute or psychologically complicated as in tech itself, where a combination of AI adjacency — the largest tech companies are all investing billions in their own AI tools, and their leaders are among the most bullish on general-purpose AI — and widely used programming assistance tools have created an apocalyptic mood among workers who are coming to believe that they might be the last generation of human software engineers (or at least the last generation for whom writing code is a good job). This, again, might well be true, or true enough to warrant worrying and even changing plans; likewise for warnings by Sam Altman and Mark Zuckerberg about AI software engineers entering the workforce this year.

Read More

7 Questions With Google Brain Founder Andrew Ng On How His Venture Studio Builds And Backs AI Startups

Crunchbase • May 21, 2025

Technology•AI•Startup Incubation•Venture Capital•Manufacturing AI

Andrew Ng, co-founder of Google Brain and Coursera, has been instrumental in advancing artificial intelligence (AI) through his venture studio, AI Fund. Established in 2018 with an initial $175 million, AI Fund focuses on building AI startups from the ground up, a departure from traditional venture capital approaches that typically invest in existing teams and ideas. (siliconangle.com)

The fund's unique model involves developing new product ideas internally and forming startups to commercialize them. This strategy allows AI Fund to maintain greater control over the direction and execution of its ventures, ensuring alignment with its vision for AI's role in society. (siliconangle.com)

One of AI Fund's notable initiatives is Landing.ai, a startup dedicated to integrating AI into the manufacturing sector. Landing.ai has developed a visual inspection system that utilizes computer vision to identify defects in products, enhancing quality control processes. This innovation exemplifies AI Fund's commitment to applying AI solutions to real-world industry challenges. (venturebeat.com)

In 2024, AI Fund launched its second venture, AI Venture Fund II, aiming to raise over $120 million. This fund continues AI Fund's mission to systematically create and support AI startups, reflecting Ng's belief in AI as "the new electricity" and its potential to transform various industries. (imaginefuture.ai)

Ng's approach emphasizes the importance of identifying concrete use cases for AI technologies and developing a repeatable process for building successful AI companies. By focusing on both the technical and business aspects of AI ventures, AI Fund seeks to bridge the gap between AI research and practical applications, fostering an ecosystem where AI can drive significant societal advancements. (swellai.com)

Read More

Claude 4 – The first universal assistant?

Exponentialview • Azeem Azhar • May 22, 2025

Technology•AI•UniversalAssistant•ToolIntegration•AnthropicClaude•Claude 4

Claude 4 – The first universal assistant?

Hi here.

The promise of a truly universal AI assistant – one that bridges thinking and doing, research and execution – is becoming a practical reality. I got early access to Anthropic’s new model, Claude 4 Opus, and had a chance to run it through its paces while preparing for annual leave. So my key question was: can Opus 4 help me get more done in one day than I could do alone or even with other LLMs?

The TL;DR is: the “open agentic web” Microsoft CTO Kevin Scott describes—a world where AI agents can act on your behalf through open, reliable, interoperable protocols—is starting to materialize. Today, I’ll walk you through five tasks I put to Opus 4—and how it handled them.

Beyond chat

Anthropic describes its new release as “our most capable hybrid model,” emphasizing improvements in coding, writing, and reasoning. The real headline, though, is its extended thinking with tools—everyday apps like email, Google Drive, and spreadsheets that the AI can open and use on your behalf. It can read your email, search for an answer, and fire back a reply—all in one sweep. Better still, you’re not limited to Anthropic-approved tools; in principle, you can hook it up to anything online. I did exactly that with Zapier, a no-code automation hub, letting Opus 4 tap into thousands of apps such as Google Sheets, Slack, and Notion.

You can enable Opus 4 to use multiple tools.

The new model promises sustained, independent work over longer periods of time. It’s built for what Anthropic calls “complex AI agent work and deep research,” staying focused through thousands of steps. This isn’t new – o3 can do this, but what Opus 4 currently promises is to integrate with a larger variety of tools alongside its superior coding ability.

To test its abilities, I threw my pre-holiday task list at Opus 4:

Checking if Azeem had assigned me any tasks.
Creating datasets from poorly formatted notes.
Brainstorming newsletter themes.
Auditing internal tools for security vulnerabilities.
Building a platformer game for my flight.

Tackling my to-do list

The Gmail integration impressed me immediately.

I simply asked it to “Check my Gmail for to-dos from Azeem in the last week.”

Opus 4 successfully scanned a week’s worth of emails, identifying every task Azeem had relayed with remarkable accuracy. It understood context, priority and the implicit requests buried in conversational emails. For example, Opus 4 identified an important testing request from Azeem where I needed to evaluate a new content transformation tool. Opus 4 didn’t just summarize the task as “test new tool.” It picked up that I needed to apply a specific template, wait for new content, compare results with an earlier version, and understand the goal: to improve content quality. It gave me the sense that I could ask it to scan anything in my inbox—and it would find it.

Research > Spreadsheets > Email

My second task was to take a casual list of links to data points I’d been meaning to turn into proper datasets, but hadn’t had the time. I thought it was a perfect way to test Opus 4’s tool use. I handed the model a messy list of URLs pointing to articles on AI companies, energy use and financial stats, and asked it to:

Review each link and identify the type of data it refers to.
Fill in any missing information through additional research.
Create a separate spreadsheet for each dataset.
Compile all spreadsheets into a final summary.
Draft an email with the compiled files to send to myself.

The results were impressive—but also revealed Opus 4’s limitations. The model parsed this chaos into six structured datasets – ranging from AI revenue tracking to the critical mineral requirements of renewables. It checked over 750 sources in the process, with each dataset including relevant metrics, time series data, and sources. These datasets weren't complete, but they're definitely a useful start.

Opus 4 managed to send the email, but only about half the data it collected made it through.

But the real test came next: “Can you create a Google Sheet for each dataset and email the sheets to my colleague?”

This is where the promise of seamless tool integration met reality. Opus 4 created the spreadsheets and drafted the email—but only the first sheet was fully populated.

It’s unclear whether the shortfall lay in Opus 4 itself or in its connection to other services. It was using Zapier—a universal adapter between tools like Google Sheets and Gmail. Opus 4 made 17 separate tool calls to populate the spreadsheet, which hints at the brittle choreography involved. Perhaps the integration wasn’t robust enough. Perhaps there were better ways to structure the task. Or perhaps this is simply what early-stage prototypes look like.

Compounding the problem, I had to restart the conversation midway through—the research step had bloated the chat’s memory, cutting off continuity. Tools like Cursor already offer workarounds for this by summarizing long threads and preserving flow, so the issue is likely fixable. But it was a reminder: even the most capable assistant is only as smooth as its connective tissue.

Comparative Intelligence

To test how different AI models handled synthesis at scale, I gave each the same chaotic input: a week’s worth of Exponential View research. The corpus spanned approximately 100,000 words—a sprawling mess of RSS-fed essays, social media fragments, internal Slack commentary, and working notes. Only Opus 4 could take in the whole thing; Gemini and o3 needed a reduced set, limited by context constraints.

The outputs didn’t just summarize—they revealed distinct thinking styles.

o3 was the clearest communicator. It surfaced seven crisply defined, executive-ready themes—from “AI goes ultra-scale” to “The subscriptionization of frontier AI” and “Agentic browsers & the birth of the Agentic Web.” Each entry combined a sharp headline, strong explanatory context, and a closing “why this matters” tailored for newsletter framing. This was newsletter-ready synthesis, built for action and attention.

Gemini took a more structured but conventional path. It grouped content by theme and evidence, producing solid, even-handed summaries that covered the major trends. But it played it safe—no strong editorial perspective, no surprising framings. It was competent coverage, but lacked edge.

Opus 4 delivered the most thought-provoking analysis. It offered system-level insights, pulling unexpected threads—like China’s “margin zero” AI strategy, or the tension between AI capabilities and national security control. Opus 4’s write-up felt more like a mini-essay: less polished, perhaps, but deeper in insight density and conceptual reach.

All three models surfaced the same macro-trends: the energy cost of scale, the rise of autonomous agents, and the lag between research and deployment. But their framing diverged—o3 took an executive-first lens, Gemini stuck to topical reporting, and Opus 4 zoomed out to system-level analysis. If o3 was a chief of staff, then Opus 4 was a policy analyst, and Gemini a competent newswire writer.

Anthropic’s wheelhouse: coding

Anthropic has always had a comparative advantage in software engineering—it’s their bread and butter. So I naturally put Opus 4 to work on a coding task. I’m not a developer by trade, but these new tools have made it easier to dabble. One recent project involved building internal AI tools for the team using Lovable, a no-code platform for generating full-stack apps. Knowing the code would likely have some rough edges, I fed it into three AI coding assistants: Claude Code (Anthropic’s coding interface powered by Opus 4), Codex and Google’s Project Jules. I asked each to identify and fix any security vulnerabilities.

Opus 4 found and fixed nine issues and appeared to be the most systematic in its approach. Codex found six, including one that Opus 4 missed—highlighting how model diversity can still pay off. Jules only found five. All three models successfully patched the flaws they identified, but Opus 4 still came out ahead—more comprehensive, more confident. Still, this was another reminder: when the stakes are high, it’s worth asking multiple models the same question and comparing what comes back.

And my last task, the mandatory fun one, was to make a platformer. Opus 4 managed to make it difficult enough that I couldn’t finish the first level (I am also just generally terrible at platformers).

The agentic web emerges

This is the first moment the “open-agentic web” we touched on earlier feels tangible, rather than theoretical. Opus 4, currently the only major chat client open to external integrations, represents early infrastructure for this vision. When it links together email scanning, data extraction, spreadsheet creation and communication, we see the beginnings of the universal assistant.

The industry is shifting fast: Google is building Gemini as a “full world model,” Apple is opening AI features to third-party developers, and OpenAI’s leaked plan is to turn ChatGPT into an “intuitive AI super-assistant.” These companies are positioning to replace traditional interfaces entirely. The “super-assistant” emerges as a new product category that tech giants are racing to define.

My early testing revealed both breakthroughs and breaking points. When Opus 4’s tool integration works, it removes friction between thinking and doing. Yet capacity timeouts, incomplete executions and integration immaturity expose the architectural challenges in building truly autonomous systems. But you still get the sense that these models are on a smooth exponential. Each new release seems more capable: smarter and better able to make use of external tools. With a bit of imagination, you can see what that might look like: near-perfect data collection, dependable email triage, and autonomous bug-fixing.

And this is the crux of the fundamental shift we’re seeing. The universal assistant arrives as an ecosystem of AI agents connected through emerging standards. We are now at the point where integration matters as much as intelligence. Opus 4’s strength lies in accessing and acting through Gmail, Drive and other tools. The question now is which model becomes the interface layer—and whether Opus 4 is the first real contender to span the full stack of tools and workflows.

Read More

Google IO

AI Overviews are now available in over 200 countries and territories, and more than 40 languages.

Blog • Hema Budaraju • May 20, 2025

Technology•AI•Search•Innovation•GlobalExpansion•Google IO

AI Overviews are now available in over 200 countries and territories, and more than 40 languages.

Google has expanded its AI Overviews feature in Search to over 200 countries and territories, making it accessible in more than 40 languages. This expansion aims to provide users worldwide with quick, AI-generated summaries for their queries. AI Overviews have been well-received, with over a billion global users each month, and are now available to teens and without the need to sign in.

In addition to the global rollout, Google has introduced AI Mode, an experimental feature that offers a conversational, chatbot-like interaction within Search. This mode leverages Google's Gemini 2.0 model to handle complex, multi-part queries, providing comprehensive responses directly in the search interface. Currently, AI Mode is available to Google One AI Premium subscribers through Labs. (aigptjournal.com)

These advancements reflect Google's commitment to integrating AI into its services, enhancing the search experience by delivering more intelligent and contextually relevant information to users worldwide.

Read More

Google finally launches NotebookLM mobile app at I/O: hands-on, first impressions

Venturebeat • May 20, 2025

Technology•AI•Productivity•MobileApp•Google•Google IO

[YOUTUBE_EMBED:VIDEO_ID]

At the 2025 Google I/O conference, Google unveiled the NotebookLM mobile app, extending its AI-powered note-taking capabilities to Android and iOS devices. This move addresses user demand for a mobile version of the previously web-only service, which allows users to upload and query documents through conversational AI.

The mobile app introduces several features tailored for on-the-go usage. Users can download Audio Overviews—AI-generated podcast discussions of uploaded documents or media—for offline playback. This functionality is particularly beneficial for multitasking and learning during commutes or travel.

The app's interface mirrors the web version, featuring tabs for organizing notebooks, including "Recent," "Shared," "Title," and "Downloaded." Each notebook entry displays a prominent play button for its "Audio Overview." Creating new notebooks is facilitated by a floating action button at the bottom. (winbuzzer.com)

Integration with the operating system’s share sheet allows users to add content like web pages or documents directly into NotebookLM from other apps. Inside a notebook, navigation relies on a bottom bar with tabs for "Sources," "Chat," and "Studio (tools)." Tablet-optimized layouts ensure a seamless experience across devices. (winbuzzer.com)

NotebookLM operates on a freemium model. The standard free tier allows up to 100 notebooks, each containing up to 50 sources (with a limit of 500,000 words per source), 50 chat queries per day, and 3 daily Audio Overview generations. For more intensive use, NotebookLM Plus offers expanded limits—500 notebooks, 300 sources each, 500 daily queries, and 20 audio generations—along with exclusive features like customizable AI personas and specific sharing options like "chat-only" notebooks. (winbuzzer.com)

The launch of the NotebookLM mobile app signifies a significant advancement in AI-driven productivity tools, offering users enhanced flexibility and functionality for managing and interacting with their notes across devices.

Read More

Our vision for building a universal AI assistant

Blog • Demis Hassabis • May 20, 2025

Technology•AI•MachineLearning•Innovation•Robotics•Google IO

Our vision for building a universal AI assistant

At Google I/O, we discussed how we're extending Gemini to become a world model. Gemini is our most advanced AI model, designed to understand and generate human-like text, images, and videos. By integrating multimodal capabilities, Gemini can process and generate content across various formats, enabling more natural and intuitive interactions.

Our vision is to build a universal AI assistant that seamlessly integrates into daily life, assisting with tasks ranging from answering questions to creating content. To achieve this, we're focusing on enhancing Gemini's contextual understanding, planning, and execution abilities. This involves improving its capacity to comprehend complex instructions, plan multi-step tasks, and execute them effectively.

A key component of this vision is Project Astra, which aims to develop a real-time, multimodal AI assistant capable of understanding and responding to visual inputs. This project will enable Gemini to interpret and interact with the physical world through images and videos, expanding its utility beyond text-based interactions.

Additionally, we're introducing Gemini Robotics, an advanced vision-language-action model that extends Gemini's capabilities into the physical realm. This model enables robots to perform a wide range of tasks by understanding and reacting to their environment, bringing AI into the physical world.

By advancing Gemini's capabilities in these areas, we aim to create an AI assistant that is not only intelligent but also adaptable and responsive to the diverse needs of users. Our commitment to responsible AI development ensures that these advancements are implemented thoughtfully, prioritizing user safety and ethical considerations.

Read More

DeepMind CEO Demis Hassabis + Google Co-Founder Sergey Brin: AGI by 2030?

Youtube • Alex Kantrowitz • May 21, 2025

Technology•AI•AGI•DeepMind•Google•Google IO

At the Google I/O developer conference, co-founder Sergey Brin and DeepMind CEO Demis Hassabis predicted that artificial general intelligence (AGI)—AI with capabilities equal to or exceeding human intelligence—could arrive by around 2030. (axios.com)

While much of the AI industry views AGI as inevitable, there is still uncertainty about its form and societal impact. Brin made an unplanned appearance during Hassabis' on-stage interview, where they discussed the requirements for achieving AGI. Hassabis emphasized that both scaling current AI models and developing new techniques will be essential, and the field may still need a few major breakthroughs. (axios.com)

Google also introduced new, less publicized AI approaches during the event, suggesting further innovative directions. Though views on AGI’s arrival vary widely within the tech community, Brin and Hassabis' projection reflects growing consensus within AI leadership circles about AGI’s approaching reality. (axios.com)

Read More

Google offers ‘AI mode’ in ‘total reimagining of search’

Ft • May 20, 2025

Technology•AI•MachineLearning•SearchInnovation•GenerativeAI•Google IO

Google has introduced "AI Mode," a significant overhaul of its search engine, enabling conversational, chatbot-like interactions similar to ChatGPT. This feature, now accessible to all U.S. users via Google Search and Chrome, is part of Google's broader initiative to integrate generative AI into its services amid growing competition from OpenAI and Anthropic. CEO Sundar Pichai described it as a "total reimagining of search," aiming to handle 8.5 billion daily queries more intelligently. (ft.com)

In addition to AI Mode, Google unveiled new AI agents, including Project Mariner, capable of complex tasks like booking travel or conducting research, and previewed Project Astra, a multimodal assistant utilizing voice and visual inputs. The company is transitioning from its traditional ad-supported model by offering paid subscriptions for advanced AI features—$25/month for "AI Pro" and $250/month for an "Ultra" package. (ft.com)

Despite privacy concerns, Google is advancing by enhancing its Gemini large language model, claiming its latest version outperforms competitors in benchmarks. The company is also implementing the Model Context Protocol to facilitate interaction between AI agents across apps and platforms, signaling its commitment to leading the evolving AI ecosystem. (ft.com)

Read More

Inside Google’s AI leap: Gemini 2.5 thinks deeper, speaks smarter and codes faster

Venturebeat • May 20, 2025

Technology•AI•MachineLearning•DeepLearning•Innovation•Google IO

Google is advancing its pursuit of a universal AI assistant with the introduction of Gemini 2.5 Pro and Flash models, featuring significant enhancements in reasoning, coding, and context understanding. These developments aim to create an AI that is personal, proactive, and powerful.

The Gemini 2.5 Pro model now includes an experimental enhanced reasoning mode called 'Deep Think,' which allows the AI to consider multiple hypotheses before responding. This feature has demonstrated impressive performance on challenging benchmarks, including the 2025 USA Mathematical Olympiad and LiveCodeBench, a difficult coding benchmark. Additionally, Gemini 2.5 Pro leads the LMArena leaderboard, which evaluates AI based on human preference.

In contrast, Gemini 2.5 Flash is optimized for speed and efficiency, making it suitable for real-time tasks such as summarizing documents, captioning images, and classifying data. It introduces 'dynamic and controllable reasoning,' allowing developers to adjust the model's processing time based on query complexity, balancing performance with cost. This flexibility is particularly beneficial for high-volume applications like customer service and real-time information processing.

Both models are available through Google AI Studio, the Gemini API, and Vertex AI, enabling developers to build and manage AI applications effectively. The integration of these advanced models signifies Google's commitment to leading the AI ecosystem by providing tools that enhance productivity and innovation across various domains.

Read More

Google I/O 2025: Everything announced at this year’s developer conference

Techcrunch • May 20, 2025

Technology•AI•MachineLearning•Android•MixedReality•Google IO

At the Google I/O 2025 developer conference, Google unveiled significant advancements across its product suite, emphasizing artificial intelligence (AI) integration. A central highlight was the introduction of Gemini Ultra, a premium subscription service priced at $249.99 per month, offering enhanced access to Google's AI-powered applications and services. Subscribers gain exclusive features, including the Veo 3 video generator, the Flow video editing app, and the advanced Gemini 2.5 Pro Deep Think mode, which enhances reasoning capabilities. Additionally, Gemini Ultra provides higher usage limits on platforms like NotebookLM and Whisk, along with access to the Gemini chatbot in Chrome, Project Mariner-powered tools, YouTube Premium, and 30TB of storage across Google Drive, Photos, and Gmail.

In AI model developments, Google introduced Deep Think within Gemini 2.5 Pro, an enhanced reasoning mode that evaluates multiple potential answers before responding, improving performance on specific benchmarks. While detailed mechanics remain undisclosed, Deep Think is currently available to "trusted testers" via the Gemini API, with a broader rollout pending further safety evaluations.

The Veo 3 video-generating AI model was also showcased, capable of producing videos complete with sound effects, background noises, and dialogue. Veo 3 offers improved footage quality over its predecessor and is accessible to Gemini Ultra subscribers through the Gemini chatbot app, where it can be prompted with text or images.

Google's AI image generation capabilities were further enhanced with the introduction of Imagen 4, which boasts faster processing speeds compared to Imagen 3. This tool is designed to generate high-quality images from textual descriptions, streamlining the creative process for users.

The conference also highlighted advancements in Android 16, featuring a new design language called Material 3 Expressive. This update aims to provide a more responsive and engaging user experience across Android devices. Additionally, Google announced the Android Show: I/O Edition, a dedicated event focusing on Android updates, including Android 16 and related platform announcements.

In the realm of mixed reality, Google previewed Android XR and Project Moohan, a collaboration with Samsung. These initiatives are expected to introduce new XR devices and software ecosystems, offering immersive digital experiences and deeper integration with Google services.

Overall, Google I/O 2025 underscored the company's commitment to integrating AI across its products, enhancing user experiences through advanced technologies and innovative services.

Read More

Demis Hassabis and Sergey Brin on AI Scaling, AGI Timeline, Robotics, Simulation Theory

Bigtechnology • May 21, 2025

Technology•AI•AGI•Robotics•SimulationTheory•Google IO

At the Google I/O developer conference, Google co-founder Sergey Brin and DeepMind CEO Demis Hassabis discussed the future of artificial intelligence (AI), emphasizing the importance of both scaling existing models and developing new techniques to achieve artificial general intelligence (AGI). Hassabis noted that while current AI models are impressive, they still lack certain capabilities, and reaching AGI may require additional breakthroughs. (axios.com)

Brin highlighted the significance of algorithmic advancements alongside computational improvements, suggesting that both are essential for progress in AI. He referenced historical examples, such as the N-body problem, where algorithmic advances have often outpaced computational ones.

The conversation also touched upon the role of scaling in AI development. Hassabis acknowledged the need for more data centers to support the growing demand for AI models like Gemini 2.5 Pro and Gemini Flash, emphasizing the importance of both scaling existing techniques and pursuing innovative breakthroughs.

Brin further emphasized the urgency of AI research, stating, "Anybody who's a computer scientist should not be retired right now. They should be working on AI."

Read More

Google I/O, The Search Funnel, Product Possibilities

Stratechery • Ben Thompson • May 21, 2025

Technology•AI•Search•GenerativeAI•ProductDevelopment•Google IO

Google I/O was impressive and overwhelming, but the only product that impressed was Search.

From Reuters:

Google said on Tuesday it would put artificial intelligence into the hands of more Web surfers while teasing a $249.99-a-month subscription for its AI power users, its latest effort to fend off growing competition from startups like OpenAI. Google unveiled the plans amid a flurry of demos that included new smart glasses during its annual I/O conference in Mountain View, California, which has adopted a tone of increased urgency since the rise of generative AI challenged the tech company’s longtime stronghold of organizing and retrieving information on the internet…

In a major update, the company said consumers across the United States now can switch Google Search into “AI Mode.” Showcased in March, opens new tab as an experiment open to test users, the feature dispenses with the Web’s standard fare in favor of computer-generated answers for complicated queries. Google also announced an “AI Ultra Plan,” which for $249.99 monthly provides users with higher limits on AI and early access to experimental tools like Project Mariner, an internet browser extension that can automate keystrokes and mouse clicks, and Deep Think, a version of its top-shelf Gemini model that is more capable of reasoning through complicated tasks…

Pichai told reporters that the rise of generative AI was not at the full expense of online search. This “feels very far from a zero-sum moment,” said Pichai. “The kind of use cases we are serving in search is dramatically expanding” because of AI.

This was, to be sure, terribly impressive: we already know that Gemini 2.5 is awesome, and the company’s Imagen 4 image model and Veo 3 video model were significant leaps forward from just a few months ago. The generative AI pillars Google dominates include algorithms, compute, and data, with a particular advantage coming from video data, especially YouTube. However, despite technological advances, there is skepticism about Google's ability to craft compelling generative AI products beyond Search, as product coherence and prioritization appear lacking.

The one clear exception at Google I/O was the focus on Search. CEO Sundar Pichai emphasized AI advancing Google's “timeless mission to organize the world’s information and make it universally accessible and useful,” with Search being the centerpiece. AI Overviews launched last year have scaled to over 1.5 billion monthly users, driving significant growth in query types and visual searches via Google Lens.

AI Mode, introduced by Head of Search Liz Reid, offers a chat-like interface powered by Gemini 2.5, designed to tackle complex questions and help users get things done. This new AI Mode is rolling out across the U.S. and is expected to integrate many of its cutting-edge features into the core search experience over time. The goal is to create a strong AI funnel where advanced features are tested in AI Mode and, once matured, "graduate" to become part of mainstream Search.

Productizing AI features involves overcoming hurdles such as performance, revenue impact, and safety, especially as Google aims to keep users from migrating entirely to competitors like ChatGPT. Usage metrics and monetization play key roles in determining which AI Mode features make it to core Search.

There is debate on how much Google’s new AI tools will disrupt startups, as Search remains its only truly functioning product. Many announced demos might not become widely used products. Devices like Android and XR glasses are key canvases for deploying AI capabilities broadly, leaving space for startups to innovate on software using AI. Google Cloud offers a big opportunity by making AI models accessible via APIs for external developers.

Overall, Google is improving Search with AI but has yet to prove it can translate technological leads into other compelling consumer products. This leaves room for startups to create meaningful new AI applications using Google’s underlying tech.

Read More

OpenAI

Why did OpenAI buy Jony Ive & Co

Om • May 22, 2025

Technology•AI•Hardware•Design•Innovation•OpenAI

OpenAI has made a significant move in the tech industry by acquiring io Products, a startup co-founded by former Apple Chief Design Officer Jony Ive, in a $6.5 billion all-stock deal. This acquisition marks OpenAI's largest to date and signifies its strategic entry into the consumer hardware market. (reuters.com)

Jony Ive, renowned for his pivotal role in designing iconic Apple products such as the iPhone, iPad, and iMac, co-founded io Products in 2024. The startup focuses on developing AI-powered devices, aiming to create "physical AI embodiments" that integrate generative AI into the physical world. These devices could range from cars and robots to AI-powered wearables, offering a new dimension to human-AI interaction. (apnews.com)

Prior to this acquisition, OpenAI held a 23% stake in io Products, reflecting a longstanding collaboration between the two entities. The full acquisition will bring io's 55 employees under OpenAI's umbrella, with Ive assuming a key creative role without becoming a formal employee. His design firm, LoveFrom, will maintain its independence while leading design efforts across OpenAI and io. (apnews.com)

The move aligns with OpenAI's vision to develop hardware tailored to the generative AI era, leveraging Ive's design expertise to potentially revolutionize devices akin to the impact of the iPhone. OpenAI CEO Sam Altman and Ive have hinted at a prototype device in development, describing it as potentially unprecedented in technological advancement. (reuters.com)

This acquisition also positions OpenAI to compete more directly with tech giants like Apple and Google in the AI hardware space. Apple's stock declined over 2% following the announcement, reflecting concerns over increased competition in the AI device market. (reuters.com)

Read More

Charted: ChatGPT’s Rising Traffic vs. Other Top Websites

Visualcapitalist • May 21, 2025

Technology•AI•ChatGPT•WebTraffic•Growth•OpenAI

Charted: ChatGPT’s Rising Traffic vs. Other Top Websites

ChatGPT has experienced significant growth in web traffic, solidifying its position as a leading AI tool. In September 2024, it surpassed 3 billion monthly visits, marking a 112% year-over-year increase. (similarweb.com) This surge has propelled ChatGPT into the ranks of the world's most visited websites, with 4.8 billion monthly visits as of November 2024, placing it ninth globally. (visualcapitalist.com)

The platform's rapid ascent is evident when compared to other AI tools. In January 2025, ChatGPT led with 4.7 billion monthly visits, followed by Canva at 887 million and Google Translate at 595 million. (visualcapitalist.com) This dominance underscores ChatGPT's widespread adoption and its pivotal role in the AI landscape.

The growth trajectory of ChatGPT has also outpaced that of traditional search engines. By September 2024, ChatGPT's traffic had surpassed that of Microsoft's Bing, which recorded 1.7 billion visits in the same month. (visualcapitalist.com) This shift highlights a growing preference for AI-driven interactions over conventional search methods.

In summary, ChatGPT's exponential growth in web traffic not only reflects its increasing popularity but also signifies a broader shift towards AI-powered platforms in the digital realm.

Read More

OpenAI's Second CEO Tasked With A Lot of First CEO Stuff

Spyglass • May 21, 2025

Technology•AI•OpenAI•Leadership•Profitability

Fidji Simo, OpenAI’s newest senior hire and former Instacart CEO, has taken on a newly created role as CEO of applications at OpenAI. She is tasked with helping the company become a profitable global business while reshaping an internal culture marked by executive infighting and high-profile departures. This position places her in charge of all major business functions, including product, finance, and sales teams, reporting directly to Sam Altman, OpenAI’s CEO.

Altman, who currently manages around two dozen direct reports, has been seeking to delegate more responsibility amid rapid company growth. He envisions a future where OpenAI’s three main divisions—applications, infrastructure, and research—each evolve into multitrillion-dollar businesses. Simo will focus on the applications side, aiming to guide OpenAI toward profitability, while Altman turns his attention more to infrastructure and research.

There is speculation about how the leadership structure may evolve. It's possible Simo becomes the overall CEO of applications, with new executives heading infrastructure and research. Altman might retain the overall CEO title or focus exclusively on research. Regardless, this marks a new generation of leadership for OpenAI, an organization that has expanded its headcount fivefold since ChatGPT’s 2022 launch and now boasts 5.1 billion monthly visits to the chatbot, ranking it just behind giants like Google, YouTube, and Facebook in web traffic.

Simo and Altman share a management style that rejects standing meetings in favor of efficiency, signaling a streamlined approach to leadership as the company prepares for a potential public offering. The move reflects OpenAI’s ambition to mature from a fast-growing startup into a structured, commercially successful enterprise.

Read More

Sam & Jony introduce io

Youtube • OpenAI • May 21, 2025

Technology•AI•Innovation•ProductLaunch•VideoContent•OpenAI

Read More

OpenAI’s next big bet won’t be a wearable: report

Techcrunch • May 22, 2025

Technology•Hardware•AI•OpenAI•Innovation

OpenAI is venturing into hardware development with a new, compact, screenless device designed to integrate seamlessly into daily life. CEO Sam Altman envisions this device as a "third core device" alongside existing technologies like the MacBook Pro and iPhone, serving as an "AI companion" that is fully aware of its user's surroundings.

This initiative follows OpenAI's recent acquisition of io, a startup founded by former Apple designer Jony Ive. The $6.5 billion equity deal aims to leverage Ive's design expertise to create innovative AI-powered hardware. Ive will assume a significant creative and design role within OpenAI, focusing on developing devices tailored for the AI era. (reuters.com)

Altman has emphasized the importance of secrecy in this project to prevent competitors from replicating the product before its official launch. The company is working diligently to develop a device that transcends traditional interfaces, offering a more integrated and intuitive user experience.

This strategic move signifies OpenAI's commitment to expanding beyond software solutions, aiming to revolutionize the way users interact with AI through innovative hardware. The collaboration with Jony Ive is expected to play a pivotal role in achieving this vision, potentially marking a significant milestone in the evolution of AI technology. (reuters.com)

Read More

Jony Ive and Sam Altman take on Apple

Ft • May 22, 2025

Technology•AI•HardwareInnovation•OpenAI•Apple

Jony Ive, the renowned former Apple designer, has joined forces with OpenAI in a landmark $6.4 billion acquisition, positioning him at the forefront of AI innovation. OpenAI acquired full ownership of Ive's AI hardware start-up io, bringing its 55 employees under its umbrella while Ive assumes a key creative role without becoming a formal employee. The move signals OpenAI CEO Sam Altman's confidence in Ive’s design vision. This partnership, potentially rivaling Apple, comes as Apple lags in AI development despite a partnership with OpenAI for Siri. Apple’s stock dropped $45 billion following the announcement.

The acquisition of io marks OpenAI's strategic push into hardware, aiming to develop devices tailored for the era of artificial general intelligence (AGI). Ive and his team will steer the design of these products, leveraging experience from developing iconic Apple products like the iPhone and iPod. Despite previous AI hardware ventures like Humane's AI pin failing commercially, OpenAI is committed to innovating beyond traditional typing-based interfaces. The move also aligns with the company's broader partnership with Apple, which recently integrated ChatGPT into its devices. Apple, too, is exploring an AI future where smartphones may become obsolete, echoing Ive’s concerns over the societal impact of these devices.

Read More

OpenAI Acquires io, OpenAI’s Strategic Positioning, Apple’s Worsening AI Problem

Stratechery • Ben Thompson • May 22, 2025

Technology•AI•OpenAI•Apple•Hardware

OpenAI is acquiring io, Jony Ive's hardware company. Is the company actually focused on displacing Apple? Apple needs to spend big in response.

On yesterday’s Sharp China, Andrew and Bill discuss the Commerce Department’s guidance on Huawei Ascend chips and the AI deal with Gulf states.

This week’s Stratechery Interview was with Nvidia CEO Jensen Huang, and was published last Monday. This coming Monday, meanwhile, is the Memorial Day holiday in the U.S. There will be no Update.

OpenAI Acquires io

From Bloomberg:

OpenAI will acquire the AI device startup co-founded by Apple Inc. veteran Jony Ive in a nearly $6.5 billion all-stock deal, joining forces with the legendary designer to make a push into hardware. The purchase — the largest in OpenAI’s history — will provide the company with a dedicated unit for developing AI-powered devices. Acquiring the secretive startup, named io, also will secure the services of Ive and other former Apple designers who were behind iconic products such as the iPhone…
For the British-born designer, the move marks a high-profile return to a consumer technology industry he helped pioneer. Working for years alongside Steve Jobs, he crafted the look and feel of the modern smartphone, in addition to the iPod, MacBook, iPad and Apple Watch. He left Apple in 2019. When Ive departed Apple, CEO Tim Cook pitched the idea that the two parties would remain collaborators. But they never released a product together after Ive’s exit. And now the designer is embarking on a new collaboration with Altman, who he called a “rare visionary.”
Ive was once described by Jobs as his “spiritual partner,” and his new stint designing rival technology products could be seen as a bad omen for Apple — a company already struggling to compete in AI. In the interview, Altman said Jobs would be “damn proud” of Ive’s latest move.
OpenAI released a nine minute and 21 second video of Ive and Altman complimenting each other, expressing love for San Francisco, and announcing the new partnership. It’s a bit much, to be honest, although beautifully shot and produced (I assume it was put together by ‘LoveFrom,’ Ive’s not-included-in-the-deal-but-not-accepting-new-clients design agency); I think the key segment from Ive’s perspective is this:
Sam Altman: We had both a very strong shared vision. We maybe didn’t know exactly where we were going to go, but the direction of the force vector felt clear. Then this deeply shared sense of values about what technology should be, when technology has been really good, when it’s gone wrong.
Jony Ive: I mean, that was in a way one of the basis, I think, one of the reasons Sam and I clicked was, despite our wonderfully different journeys to this point, our motivations and values are completely the same. In my experience, if you’re trying to have a sense of where you are going to end up, you shouldn’t look at the technology, you should look at the people who are making the decisions, and you should look at what drives, motivates and look at values.
Ive expressed what drives and motivates him earlier this month in this excellent interview with Patrick Collison (which John Gruber and I discussed on Dithering):
You’re talking a lot about the purpose of design and the effect that design has on the recipient, on the user, on the consumer, whatever the case is. There’s widespread concern and speculation about the effects of smartphones and the Internet, doesn’t necessarily accord just with the smartphone, but on some of these products on attention spans, and whether it has some adverse effect on kids or teens or who knows, maybe all of us, maybe the adults as well. There’s questions over with AI, whether it changes how education works, and cheating, and school. All of these technologies that we create have this potential double-sidedness to them, and so I guess as somebody who clearly takes seriously and thinks seriously about the full effects, how do you think about the possible harms?
Jony Ive: Yeah, I think [there is] probably not anything that I can be more preoccupied or bothered by than what you just described. I think when you’re innovating of course there will be unintended consequences. You hope that the majority will be pleasant surprises. Certain products that I’ve been very, very involved with, I think there were some unintended consequences that were far from pleasant. My issue is that even though there was no intention, I think there still needs to be responsibility, and that weighs on me, as you know heavily.

It seems pretty clear that Ive is talking about the iPhone, which is to say it sure seems like he is motivated not simply to build an AI device, but to actually diminish the iPhone’s dominance in users’ lives, or even, in the long run, kill it completely. This tracks with the Wall Street Journal’s reporting on their plans:

OpenAI Chief Executive Officer Sam Altman and Ive’s design firm, LoveFrom, have been working on a new device that will move consumers beyond screens, according to people familiar with the matter. They have been collaborating for two years on a closely guarded project, considering options including headphones and other devices with cameras, the people said.

That’s admittedly pretty thin gruel, but it’s in-line with other rumors, and, more importantly from my perspective, tracks with Ive’s expressed sentiment. And, it should be noted, this is a sentiment that Altman has expressed as well. Here’s an example from before OpenAI even existed:

people sacrifice actual happiness and actual accomplishment for short-term dopamine hits by posting and chasing likes/RTs on FB/Twitter

Here’s another a year after OpenAI was founded:

Digital addiction is going to be one of the great mental health crises of our time.
Altman wrote on his blog in 2017:
I believe attention hacking is going to be the sugar epidemic of this generation. I can feel the changes in my own life — I can still wistfully remember when I had an attention span. My friends’ young children don’t even know that’s something they should miss. I am angry and unhappy more often, but I channel it into productive change less often, instead chasing the dual dopamine hits of likes and outrage.
It’s always challenging with Altman to know when or if he is talking his book, as it were, but there is a record of long-running sentiments that might convince Ive that their “motivations and values are completely the same”; at the same time, there are certainly other motivations as well. Ive, for his part, just became a lot richer, at least on paper; Altman may or may not have diluted the non-profit’s share of OpenAI by making a big purchase with stock. He also is sliding into the role occupied by his childhood idol, one Steve Jobs.

OpenAI’s Strategic Positioning

The larger question is what this means for OpenAI. On one hand, the angle here is obvious, and fortuitously articulated by me just yesterday:

It has long been the case that the best way to bring products to the consumer market is via devices, and that seems truer than ever: Android is probably going to be the most important canvas for shipping a lot of these capabilities, and Google’s XR glasses were pretty compelling (and, in my opinion, had a UX much closer to what I envision for XR than Meta’s Orion did). Devices drive usage at scale, but that actually leaves a lot of room for startups to build software products that incorporate AI to solve problems that people didn’t know they had; the challenge will be in reaching them, which is to say the startup problem is the same as ever.

….

Read More

Space

Has Starlink already won the new space race?

Ft • May 22, 2025

Technology•Space•SatelliteInternet•GlobalConnectivity•Geopolitics

Elon Musk's Starlink satellite system currently leads the burgeoning low Earth orbit (LEO) satellite broadband market, boasting over 7,300 active satellites and serving 5 million users across 125 countries. This dominance is underpinned by frequent, low-cost launches, rapid innovation, and a focus on consumer-friendly, affordable terminals. Starlink aims to expand its constellation to 40,000 satellites and projects $12 billion in revenue for 2025.

Amazon's Project Kuiper, with its recent satellite deployment, poses the most significant Western challenge, intending to leverage Amazon's global retail and cloud assets. However, Kuiper faces production delays and high costs, requiring up to $20 billion in investment. Other contenders include Eutelsat's OneWeb, Telesat’s Lightspeed, and Chinese state-supported constellations Guowang and SpaceSail, which are rapidly advancing despite launch capacity constraints.

The geopolitical landscape is also heating up, with countries like China and the EU pursuing satellite autonomy to counter US dominance. China’s integration of LEO networks with its Belt and Road Initiative adds a strategic layer to this new space race. Regulatory shifts, price wars, and the push for sovereignty in satellite infrastructure are shaping what is not just a commercial battle but a geopolitical contest for global digital influence.

Read More

From SignalRank

The problem with interval funds

Signalrankupdate • May 19, 2025

Finance•Investment•PrivateMarkets•AlternativeInvestments•IntervalFunds•From SignalRank

Interval funds, a financial product that offers capped liquidity at fixed intervals, have seen a resurgence in popularity, especially in private markets such as venture capital (VC) and private equity (PE). Originally invented in the 1990s, these structures have grown to a combined net asset value of over $360 billion across real estate, credit, and private markets. According to Goldman Sachs, these semi-liquid funds have experienced a compound annual growth rate (CAGR) of around 60% from 2021 to 2025, driven largely by private credit and now expanding into retail spaces with VC and growth equity.

The attraction of interval funds lies in their promise to democratize access to traditionally illiquid and exclusive alternatives by lowering minimum investment thresholds and offering greater liquidity than conventional GP/LP private funds. Fund managers such as Hamilton Lane, Stepstone, Coatue, and Ark Ventures have launched products targeting retail investors with this model, aiming to provide periodic redemption opportunities while maintaining exposure to private assets.

However, the key issue with interval funds, particularly in illiquid PE/VC contexts, is asset-liability mismatching. Managers often allocate only a minority portion (sometimes as low as 20-50%) of the fund into actual private assets, with the substantial remainder held in liquid public stocks and cash to meet redemption demands. This approach can dilute the private asset exposure that investors primarily seek. For instance, Coatue’s interval fund reportedly invests a minority in private investments, while the bulk of the fund remains in public equities and cash equivalents.

This setup creates a layered investment risk where investors must bet simultaneously on the manager’s skill in private markets, public markets, and asset allocation timing. It arguably places investors off the efficient frontier of investment performance, as the hybrid structure does not optimally serve either the liquidity or the private asset access goals. The article suggests an alternative model: investors controlling their public stock and cash portions independently, while using a listed closed-end fund to hold 100% private stock exposure. This model allows liquidity through share trading without forcing underlying asset liquidation and aligns better with investor interests.

The renewed interest in interval funds is partly driven by industry trends positioning semi-liquid products as a way to broaden retail investor participation in alternatives. Optimists argue this democratization reduces gatekeeper dominance, increases transparency, and channels more capital into private markets. Conversely, cynics warn that institutional investors might be offloading less desirable 2020-21 vintage assets onto retail investors under this structure, referencing Bloomberg’s Matt Levine’s critique of retail investors paying full price for assets institutions buy at substantial discounts.

The article underscores that interval funds, especially those marketed as “hybrid” solutions combining private and public assets, likely fail to meet their promises of liquidity and exposure simultaneously. These hybrid products face parallels with other "hybrid" concepts (like hybrid cars or animals), where the combined benefits are often compromised.

Looking ahead, the private markets are expected to evolve towards more public market-like structures, including liquid, low-cost passive index funds that provide full liquidity. Given the challenge of consistently outperforming public benchmarks, a shift toward index-like vehicles in private markets may gain traction, challenging traditional active management. Meanwhile, retail and sovereign investors will continue to drive the growth of alternative investments, requiring managers to adapt strategies to cater to this expanding client base. Leading firms like Blackstone and KKR are poised to capitalize on this trend as private equity and venture capital become more accessible to the broader market.

In summary, interval funds represent an interesting but flawed approach to providing liquidity and access to private markets. Their structural limitations dilute investment theses and pose complex risks for investors. The future may lie instead in clearer, more transparent vehicles with better liquidity management, aligning with broader market trends towards indexation and retail participation in alternatives.

Read More

Government Overreach

An Epic Game

Spyglass • M.G. Siegler • May 21, 2025

Technology•Gaming•AppStore•EpicGames•Fortnite•Government Overreach

Apple finally read the room. The courtroom, at least. Here's Chance Miller with 9to5Mac:

After a nearly five-year hiatus, Fortnite is back on the App Store for iPhone and iPad users in the United States. Epic Games announced the return of the battle royale gaming app this afternoon, and you can head to the App Store now to download it. Fortnite is also back in the Epic Games Store and AltStore in the European Union.

It is both wild that it has been nearly five years since Fortnite left the App Store – but far more wild that Epic has kept up the battle this long, despite what must be billions in lost revenue. It's easy to get lost in the day-to-day of this back-and-forth, but just take a step back: Fortnite is one of, if not the, biggest games in the world. And it wasn’t on the largest platform for half a decade.

And now it’s back. I would argue – as I have been for the past five years – that it’s the direct result of a calculated long-game maneuver by Tim Sweeney. Per above, it could not have been worth it, monetarily, obviously. Even with the ability to accept payments on the web, saving Epic from Apple’s 30% cut, it will take years to earn back that lost revenue – if they ever do, because you have to assume most customers will still choose to use Apple’s in-app payment mechanisms!

That’s one of the silliest things about all of this. Had Apple just agreed to compete for the customer’s wallet here, they undoubtedly would have won most of the time – yes, even with the 30% cut. Because convenience often trumps cost, and Apple’s system is seamless and very well done!

But we’ve been over all of that, again over the past many years. Apple has now changed their ways only because they were forced to. The question now is if any of this will stand, long term. Apple is, of course, appealing the Judge’s ruling. If they win, do they dare pull all of this back? We already have some of the biggest companies and apps now taking advantage of the web payments, from Amazon to Spotify. What do they do if Apple wins and puts the genie back in the bottle? Just quietly comply and pull back their web payment links? Spotify?!

And if they get loud again, what do consumers do? The answer there is still probably nothing. Their only real option is to move to Android or to Europe – I’m not sure which is easier. But Google is working like hell to make the former much compelling by the day thanks to their prowess in a space where Apple has a distinct weakness: AI. We’ll see.

To go back to Sweeney, this is now the second time he’s called Apple’s bluff – and won. That’s rather incredible. And it was nearly the same situation. Last March, Apple had to backtrack from revoking Epic’s EU developer license in the face of threats from the EU. There, as here, Apple probably had good legal footing for their move since Epic did violate their rules to get banned in the first place. But that’s not the battle Apple was actually fighting. This was a PR war being waged by Sweeney, and he was slowly but surely winning it.

Now he’s doing so less slowly but more surely. As I wrote last weekend, thinking through Sweeney’s strategy here:

If I’m him, here’s the general game plan:

1) Re-submit Fortnite to the US App Store even though you have no legal grounds to do so. No one will care about that. They will have just read about your legal win and assume you won everything and so Fortnite can return – even though this particular aspect of the case had nothing to do with that.

2) When Apple rejects (or refuses to rule) on the new submission, pull your app around the world under the notion that the unified apps all have to be updated in unison, including an element bringing the US back to the App Store. So yeah, blame Apple for this. It may even technically be true, but it doesn’t matter. Again, it’s a perception thing.

3) File a new legal claim against Apple for blocking your submission in light of the recent ruling. Again, this has no legal grounds, but perhaps the Judge who issued that ruling is, in fact, pissed off enough to entertain this in some way – even if just in weighing in on it to dismiss it sympathetically, thus generating more press, instead of immediately dismissing it, legally.

4) Give more interviews about all of the above in the coming weeks. Again, leading up to WWDC. Keep the pressure on.

If I’m right, step three may happen this week, ….

¹ Sure, some of it might be indirect -- i.e. beyond the direct revenue lost from iOS players, there’s also revenue Fortnite would have made if friends could play with other friends who have iOS devices, but because they couldn’t...

² They’ll have their own fees to pay, of course – notably, card processing fees, but those are more like 3%-4%.

³ Because there’s no way they’re going to be okay with Apple’s silly "27%" cut, which equated to more than 30% when you layer in doing your own credit card processing.

Read More

Startup of the Week

Open social web browser Surf makes it easier for anyone to build custom feeds

Techcrunch • Sarah Perez • May 22, 2025

Technology•Software•SocialMedia•Decentralization•ContentCuration•Startup of the Week

Surf, Flipboard's new app, is revolutionizing the way users interact with the open social web by enabling the creation of custom feeds that aggregate content from decentralized platforms like Mastodon, Bluesky, Threads, RSS feeds, podcasts, and YouTube videos. This approach empowers individuals to curate their own content streams, focusing on topics that genuinely interest them, such as hobbies, sports, or specific communities.

The introduction of Starter Sets simplifies the process of building and personalizing these custom feeds. Organized around popular categories, Starter Sets come pre-populated with recommended sources, allowing users to effortlessly create feeds tailored to their interests. For example, selecting a "Hobbies" feed provides access to subtopics like cycling, gaming, baking, and more, with the option to add personal social account feeds from Mastodon or Bluesky, filtered by the chosen topic. Users can also utilize the search bar to incorporate specific sources and apply filters to ensure the content aligns with their feed's focus.

By offering these tools, Surf aims to give users greater control over their social media experience, moving away from algorithm-driven timelines and fostering a more personalized and decentralized content discovery process.

Google CEO Admits Biggest Regret: Not Buying Netflix When He Had the Chance!

Youtube • All-In Podcast • May 16, 2025

Technology•Business•StreamingServices•Acquisitions•StrategicDecisions•Post of the Week

Google CEO Sundar Pichai has expressed significant regret over not acquiring Netflix during his tenure. This admission was made during an episode of the All-In Podcast, where Pichai discussed pivotal moments and decisions in his leadership journey. He noted that Google had "super intensely" debated buying Netflix, highlighting the complexity of corporate decision-making and the potential butterfly effects of such choices.

Pichai's acknowledgment of this missed opportunity underscores the strategic importance of content ownership and distribution in the digital entertainment landscape. Netflix, founded in 1997, has grown to become a dominant force in the streaming industry, boasting over 230 million subscribers worldwide as of 2025. Its success is attributed to its vast content library, original programming, and global reach.

An acquisition by Google could have significantly bolstered the tech giant's presence in the entertainment sector, complementing its existing platforms like YouTube and Google TV. Industry analysts speculate that integrating Netflix's streaming capabilities with Google's technological infrastructure and advertising prowess could have created a formidable competitor to other tech and media conglomerates.

This revelation also opens discussions about other strategic decisions and missed opportunities within the tech industry. For instance, former Yahoo CEO Marissa Mayer once admitted that acquiring Netflix or Hulu would have been a better choice than purchasing Tumblr, reflecting on the long-term value and growth potential of streaming platforms.

As the digital entertainment landscape continues to evolve, tech giants are increasingly recognizing the value of content ownership and distribution. Google's experiences and reflections, as shared by Pichai, may influence future strategic decisions, emphasizing the need for proactive engagement in emerging sectors.

Read More

A reminder for new readers. Each week, That Was The Week, includes a collection of selected essays on critical issues in tech, startups, and venture capital.

I choose the articles based on their interest to me. The selections often include viewpoints I can't entirely agree with. I include them if they make me think or add to my knowledge. Click on the headline, the contents section link, or the ‘Read More’ link at the bottom of each piece to go to the original.

I express my point of view in the editorial and the weekly video.