0:00
/
0:00
Transcript

Who's Cheating?

And who's changing?

Contents


Editorial: Who’s Cheating?

When I sit down to write my weekly editorial I try to see some kind of passion subject (for me) in the collection. Most weeks a pattern emerges. This week it is about cheating.

The New York Magazine published a piece by James D. Walsh titled “Everybody is Cheating their way through College.” It was inspired by the story of Chungin “Roy” Lee from Columbia University after he was found to have used AI to complete 80% plus of his projects. He was expelled but has not launched an app - Cluely. The app promotes itself as ‘Invisible AI to cheat on everything’.

He was pushed to do so after academia flatly refused to accept that AI can be used as a tool for academic progress. He was a cheat!

I might be wrong but I think Columbia is wrong and Lee may well have the last laugh.

Lets stand back to 60,000 feet and consider society’s options. in the past six months we have gained access to significant tools that have a breadth and depth of training on the world’s knowledge. Yes, six months. before that the early generation of AI was inclided to make things up. It still does, a little. But way less. And even that can be avoided by using the better reasoning models.

LLMs are not AGI. But they are very good.

So how should education systems react to the fact that anybody that can afford a smartphone (5 billion people in case you are wondering) has access to anything in the field of human knowledge? And it is not like a big library where you have to find the sources and read them. AI has already done that. You can ask it for context, facts, thoughts, answers to questions, references….you name it.

Why would education any longer consist of finding facts, learning them and repeating them with some kind of structure, maybe an essay?

Why would it be cheating to use this new found capability? Imagine if we said it was cheating to get a plane and we had to travel by car. Or cheating to use a computer, we have to use a typewriter.

What has to change is the expectations of the student. now that we have AI the bar should be raised, and those who use it to the best advantage should score higher than those who simply copy and paste its answers to mundane prompts.

Student capability can be boosted by a significant uplift, and marking students can take that into account. The expulsion of Lee was an admission that the systems are behind the technology. It is these systems that are in error, trying to hold back education to a 19th century standard. If they do not I predict new educational institutions not yet born will do so.

If Lee is a bad student then this approach would penalize him through his performance mark. If he is a good student seeking modern tools it would reward him.

Google was found cheating by the courts and this week speculation is rife that it is challenged on three fronts.

  1. Apple’s Eddy Cue said that Safari may become and AI browser. Google’s stock slumped 8% on the news.

  2. The courts indicated they want Google top sell its advertising exchange - which the company said is not technically feasible.

  3. OpenAI was joined by Anthropic on building web search into its APIs for developers.

Casey Newton’s essay ‘Stats from a dying web’ discusses the possible implications. Newton associates Google’s health with the health of the web and publishing and he mourns its possible eclipse by AI discovery:

For more than a year now, I’ve had one eye trained on how generative AI will reshape the web. My primary fear has been that large language models like those found in ChatGPT are now good enough that large numbers of people are beginning to abandon traditional search engines, starving publishers and websites of the traffic and money they need to continue operating.

and then:

For the moment, Google’s AI overviews seem to have quelled the possibility of a sudden mass defection away from its core search engine. But it’s now clear that for the first time in decades, a generation is growing up with the possibility of using something other than Google as its default search. Today, they’re using ChatGPT to do all their homework assignments. By the time they graduate, they may be using it to do almost everything else.

This observation by Casey is not wrong, but neither is it best viewed by asking who is threatened by it (publishers who depend on traditional web search). Any more than teachers who do not want to change are the right prism for understanding the benefits of AI in education.

AI is an opportunity for teachers, and for publishers, and for Google. But the innovators dilemma affects each of them. The fear that evokes should be a source of inspiration and passion to change and get ahead of the parade.

Essays

AI May Bring Civilizational-scale Change & The New Enlightenment

Peterleyden • Peter Leyden • May 6, 2025

Technology•AI•CivilizationalChange•FutureOfWork•NewEnlightenment•Essays

AI May Bring Civilizational-scale Change & The New Enlightenment

Some people think that artificial intelligence is over-hyped. They point to how little evidence we see of it having an impact on the economy or society and the general lack of change in the nature of work to date.

I think that artificial intelligence is actually under-hyped and that most people are not thinking nearly big enough about the scale of changes that will be coming in the next 25 years. I think we have to start thinking about the probability that civilizational-scale change lies just ahead of us.

The arrival of intelligent machines is a world-historic development that will give humans such a tremendous step-change in our capabilities that we will be able to fundamentally rework pretty much all aspects of our civilization.

As many are currently speculating, we will be able to reinvent the economy, society, and the nature of work. (Just give it more than two years, folks.) But we will also be able to rework many of our core institutions, including how we govern ourselves in democracies and how we coordinate between different political entities around the planet.

Artificial intelligence will give humans super-tools that dramatically expand what we currently know about ourselves, the world, the planet, and the universe. Human minds augmented with artificial minds will be able to see and understand things that we had absolutely no idea about until now.

AI will accelerate the development and use of other world-historic technologies, such as clean energy and bioengineering, that will impact us on a civilization scale, too. Being able to engineer all living things, from viruses to animals to ourselves, will set us apart from every other civilization in world history. The ability to take advantage of vast amounts of clean energy will also make us unique.

And, of course, in no other civilization have humans lived and worked side-by-side with intelligent machines.

The last time humans found themselves in such a fortuitous moment in history — the last time we created a new kind of civilization — was during The Enlightenment, the era from which the modern world was born.

We can see many striking parallels between what seems to be unfolding around us today and what happened during The Enlightenment, which took place in Western Europe and the United States from about 1680 to 1800.

That era witnessed the arrival of its own transformative new technology: the mechanical steam engine, which dramatically augmented the physical power of humans. That era also had a breakthrough in energy with coal mining at scale.

During The Enlightenment, the invention of financial capitalism and industrial production fundamentally reworked the economy. The creation of institutions for representative democracy gave us a new way to govern ourselves, while nation states gave us a new way to coordinate with entities around the planet.

The name “The Enlightenment” itself refers to the dramatic expansion of humans’ understanding of the world during this period. This increase in knowledge was partly enabled by new tools, like microscopes and telescopes. The people of that time thought of themselves as shining a light into fields that had previously been dark due to our ignorance, superstitions, or religious traditions.

I think we’re in a very similar moment today. So, to help us prepare for the civilization-scale change that looms on the horizon, I’m going to point out the many parallels between our era and the last time humans went through such a change.

We might be in the relatively early stages of what we someday may call “The New Enlightenment.” Hear me out.

All inventions are not equal. Some are such game-changers that they deserve to be put in a different league. Not all inventions are technologies or tools, either. Some of humanities’ most important inventions could be categorized as systems or institutions.

I use the term “mega-inventions” to set these breakthrough inventions apart from the rest, and the innovators of The Enlightenment came up with six of them...

These six mega-inventions changed civilization in profound ways, and the world we live in today is still rooted in all of them. However, after 250 years, they are getting long in the tooth and, in many cases, causing more problems than they are solving:

Mechanical engines: Throughout every civilization that came before The Enlightenment, humans’ primary sources of power were their own physical muscles and the muscles of any animals they could domesticate. This put serious constraints on the amount of work they could do. But then came the mechanical steam engine. This game-changer invention could deliver the power of 20 horses (and eventually far more), which drove an enormous amount of productivity gains and economic growth. We would not have our prosperous modern world without it.

….

The next 25 years could be a wild period driving the emergence of The New Enlightenment.

Read More

Stats from a dying web

Platformer • Casey Newton • May 8, 2025

Technology•Web•AI•SearchEngines•DigitalMedia•Essays•Essays

Stats from a dying web

The article explores the shifting landscape of web search and the broader impact of generative AI on internet usage, journalism, and information discovery. A significant event marking this change was the revelation by Apple’s Eddy Cue during Google’s antitrust trial that, for the first time in 22 years, Google searches on Safari had declined. Cue attributed this drop to the rising use of AI-driven search alternatives like ChatGPT, Perplexity, Gemini, and Microsoft Copilot. This decline caused Google’s stock to momentarily drop by 7.5%, reflecting investor fears about AI’s disruptive influence on traditional search dominance.

Google responded by confirming the accuracy of Cue’s observation but reassured the public by stating that overall search queries continue to grow, especially across Apple devices and other platforms like the Google app, voice search, and Google Lens. This suggests a channel shift rather than a total loss of users. Supporting this view, Rand Fishkin, a search optimization expert, reported that Google searches actually grew by 21.64% between 2023 and 2024. However, despite this growth in search volume, the introduction of AI overviews by Google has led to a steep 70-80% decline in click-through rates to web pages. This reduced traffic to websites is seriously hurting web-based publishers and has contributed to layoffs across major news organizations such as CNN, Vox Media, HuffPost, and NBC.

The article highlights the existential threat AI poses to the traditional web ecosystem. While Google tries to frame this as an evolution in search behavior, the consequence is a diminished role for publishers who rely on search referrals for revenue. The nature of information retrieval is changing, with users increasingly relying on AI to provide direct answers rather than clicking through to sources, which cuts deeply into the economics of online content creation.

The article also discusses the broader cultural shift AI is causing, particularly in education. A New York Magazine piece revealed widespread use of AI by college students to complete homework assignments, with usage jumping from 66% in 2024 to 92% in 2025. Students are using multiple AI platforms including Google’s Gemini and Anthropic’s Claude, not just ChatGPT. This widespread adoption signals a generational change in how knowledge and problem-solving are approached, moving away from traditional research methods to AI-powered assistance. Consequently, Google is aggressively pushing Gemini, trying to maintain relevance by pre-installing it on smartphones.

In summary, the piece outlines a transition phase where traditional web search and journalism face disruption due to AI’s rapid adoption. The impact is twofold: a decline in direct Google searches on some platforms like Safari, and a dramatic reduction in traffic to web pages caused by AI-generated summaries. This jeopardizes the financial viability of many digital content producers. The rise of AI in education and search signals changing habits that could permanently alter the web’s structure and the dominance of legacy search engines. Google’s efforts to retain users through new AI features and strategic partnerships underscore the competitive and existential pressures it faces in this evolving landscape.

Key Takeaways:

  • Apple's Eddy Cue confirmed a historic decline in Google search usage via Safari, linking it to the growing popularity of AI search tools.

  • Google maintains total search volume is up, but AI-driven answers reduce traffic to websites by 70-80%, threatening publishers and journalists.

  • Major news outlets have reported layoffs as digital ad revenue shrinks due to fewer click-throughs.

  • College students’ use of AI for schoolwork has surged, indicating a new norm in education and information consumption.

  • Google is actively promoting its AI-powered chatbot Gemini to compete with ChatGPT and others, aiming to maintain its dominant position.

  • The web’s future sees a shift away from traditional search and web browsing towards AI-driven chat and direct answers.

Stats from a dying web

Read More

Meta Earnings, Meta’s Deteriorating Ad Metrics, CapEx Meta

Stratechery • Ben Thompson • May 6, 2025

Technology•AI•Advertising•CapitalExpenditure•MetaPlatforms•Essays

Meta Earnings, Meta's Deteriorating Ad Metrics, CapEx Meta

Meta's earnings were received positively by investors, but there were definite causes for concern.

From the Wall Street Journal:

Meta Platforms said sales reached $42 billion in the first quarter and indicated growth would remain steady in the coming months, blunting concerns that President Trump’s tariffs would harm its global digital ads business. The social-media giant said its sales grew by 16%, ahead of analyst expectations. Net income for the January-to-March period was $16.6 billion. Meta, which operates apps including Facebook, Instagram and WhatsApp, projected revenue for the current quarter would increase between 8% and 16% year-over-year, sending shares up more than 3% in after-hours trading.

“I think we’re well positioned to weather the macroeconomic uncertainty,” Meta Chief Executive Mark Zuckerberg said on a call with investors and analysts Wednesday. The company said its capital spending for the year would rise from a range of $60 billion to $65 billion to $64 billion to $72 billion, largely on its artificial-intelligence investments. Overall expenses are set to fall by about $1 billion.

What accentuated this feeling for me is that I personally felt a bit more unease, both from the earnings themselves and also from the meta perspective of my interview with Zuckerberg.

Meta’s Deteriorating Ad Metrics

The first thing to note is the leveling out of impression growth even as Daily Active Person’s growth increased slightly; what this means is that the Reels inventory growth tailwind may be over. That’s always a bearish signal, as noted a year ago in Meta and Reasonable Doubt:

The most optimistic time for Meta’s advertising business is, counter-intuitively, when the price-per-ad is dropping, because that means that impressions are increasing. This means that Meta is creating new long-term revenue opportunities, even as its ads become cost competitive with more of its competitors; it’s also notable that this is the point when previous investor freak-outs have happened.

Notice, though, that this time is different; CFO Susan Li said on the earnings call:

One thing I’d share, for example, is that we actually grew conversions at a faster rate than we grew impressions over the course of this quarter. So we are — we’re expecting to — which basically suggests that our conversion grade is growing and is one of the ways in which our ads are becoming more performant.

This is true, but it also means this specific moment in time is much less bullish for Meta’s advertising business than past stock drops: impressions growth declining means that price-per-ad is the primary route for revenue growth, which will happen but will also open the door for more of Meta’s competitors. Yes, those future advertising opportunities that Zuckerberg talked about will probably lead to another inversion at some point, but not only are those opportunities uncertain, but they also are quite a ways in the future (and the bill for GPUs is today).

There is some small amount of hope on the horizon, now that Meta is starting to monetize Threads, but any material impact is still a ways out; Li said on the earnings call:

We continue to optimize ad supply across each surface to better deliver ads at the time and place they are most relevant to people. We are also starting to introduce ads on unmonetized surfaces, like Threads, which we opened up to all eligible advertisers this month to reach people in over 30 different markets to start, including the US. As we do for any newly monetized surface, we expect to gradually ramp ad supply as we optimize the ad formats and ensure they feel native to the app. We don’t expect Threads to be a meaningful driver of overall impression or revenue growth in 2025.

Zuckerberg disclosed that Threads is up to 350 million monthly active users, and said that improvements to Meta’s recommendation algorithms — including using LLMs on text-heavy Threads content — increased engagement by 35% over the last six months. Still, these MAU numbers — including the fact they are reporting MAUs, not DAUs — are relatively small potatoes, which is why the benefits are at least a year away, and probably pretty small at that.

Secondly, and more concerningly, is that while price-per-ad growth outpaced impression growth, it also decreased quarter-over-quarter. One of the reasons why 2023 was such a good time to own Meta is that a combination of figuring out ATT and expanding Reels inventory meant that both impressions and price-per-ad increased simultaneously; in contrast, it is a lot less attractive when both are decreasing simultaneously.

Read More

Philosophy Mondays: Values (Part 3)

Paragraph • continuations@newsletter.paragraph.com (Albert Wenger) • May 5, 2025

Culture•Philosophy•Ethics•Values•Knowledge•Essays


In the last two Philosophy Mondays we first defined what values are and then posed the question of a basis for values. The key requirements for such a basis is that it has to be objective so that it can be universal, i.e. apply to all humans. This will be the subject of today's post. Let's first consider some possibilities that fall short of this goal.

What about a God? Much as religions would like to make this claim, there is nothing objective about the existence of a God. That could of course change if one manifested in the present or future. It's an interesting question to speculate about what we would need to see from such a manifestation to consider them a God. Maybe that's for another post. For today's purposes it suffices to say that all the evidence brought forth to date about the existence of gods has a highly narrative quality to it or is an outright (and known) invention. One big give away here is that these are all regional in origin and have made claims to universality only over time (and often through conquest). This critique echoes Ludwig Feuerbach's argument that gods are projections of human ideals rather than objective realities.

What about consciousness? If only we had a handle on what that even is. Yes there are lots of theories floating around about consciousness but it feels only slightly less elusive than a God. It is absolutely fun to read the theories and speculate oneself about what consciousness might be. Of course I would advise to first try to define a bit better what one is even talking about in the first place. For now though it really won't do as a basis for values, with theories as disparate as consciousness being a fundamental aspect of all matter (as in panpsychism accepted by philosophers like David Chalmers) to consciousness being an emergent phenomenon that only occurs in complex systems (and possibly even requires language as a precondition, as suggested by thinkers like Daniel Dennett).

What about emotions? These are a pretty objective fact of our existence. These days we even understand their neural correlates in the limbic system, as well as their relationship to homeostasis and other bodily functions. They might make a potentially good starting point but there is one problem: we share them with all mammals to varying degrees. Now one might consider this a feature and not a bug. After all if emotions are the basis for values then the values would maybe apply to animals as well as to us. But this immediately runs into several problems: the biggest one of these is that the animal world is full of extraordinary violence – despite being around much longer than humans, animals haven't developed values. So we are left returning to something that is in fact germane to humans. Martha Nussbaum has provided a cogent critique of purely sentiment-based ethical theories.

Read More

R.I.P. Skype!

Om • Om Malik • May 6, 2025

Technology•Software•Communication•Skype•InternetEvolution•Essays


On May 5, Microsoft retired Skype, the startup that sparked a communication revolution. I won’t repeat myself, as I’ve already published a postmortem analyzing Skype’s decline under Microsoft’s 14-year ownership of the platform, for which it paid $8.5 billion.

Just like Nokia, Skype created one of the most iconic internet ringtones, and it will likely exist in archives. At its peak in 2009, Skype had 405 million users. They all probably heard it. To me, it will always represent what the internet sounded like in the 2000s.

Having covered Skype since its early days, it’s striking how the platform predicted nearly everything about modern digital communication but ultimately became irrelevant amidst the greatest communication boom.

Skype’s peer-to-peer architecture, adapted from Kazaa, proved to be both its greatest asset and fatal flaw. While revolutionary when introduced, the system struggled to adapt to mobile computing, where battery life and constant connectivity demanded different approaches. The specialized infrastructure became increasingly difficult to maintain and modernize. As the technology landscape evolved, Skype faced additional challenges. The platform’s unique structure collided with Microsoft’s corporate ambitions following its acquisition in 2011. Moreover, the infusion of new management mediocrity.

Skype pioneered a unified approach to communication – combining voice, video, and messaging – that created a blueprint modern platforms still follow, even as they’ve surpassed their predecessor. WhatsApp stands as the prime example of Skype’s integrated communication model. While Skype laid the groundwork, WhatsApp has taken the concept further, becoming the world’s most popular messaging app with over 2 billion users. Its success stems from its simplicity and accessibility, offering a seamless experience across devices and platforms.

The shutdown offers an opportunity to reflect on how the internet has evolved. After contemplating the Skype shutdown and observing other changes, I’ve reached a simple, obvious conclusion: the old internet is dying. More on that soon.

The evolution from Skype’s decentralized peer-to-peer model to today’s platform-centric approach reflects a broader transformation of the internet. Users have exchanged the democratic ideals of early internet protocols for the convenience of centralized services. While these newer platforms offer technical advantages, the shift raises questions about innovation and user autonomy, as illustrated by the contrast between Skype and Meta-owned WhatsApp. Looking forward, Skype’s legacy raises an important question: In our platform-dominated future, is there still room for revolutionary communication technologies that prioritize user independence over corporate control?

The answer may determine whether we’ll see another Skype-like innovation emerge from the tech industry’s margins. While the decentralized internet (Web3) could potentially steer the open web in a new direction, it remains in the grip of crypto-grifters.

Skype, R.I.P. — you were an amazing, joyful part of my internet journey. Even my mom would agree!

Read More

How To Make Money in Venture | Josh Kopelman, Co-Founder of First Round Capital

Youtube • Uncapped with Jack Altman • May 1, 2025

Business•VentureCapital•Investment•Startups•Entrepreneurship•Essays


In this insightful conversation, Josh Kopelman, co-founder of First Round Capital, shares his expertise on making money in venture capital. The discussion covers key aspects of successful venture investing, portfolio management, and the evolution of the venture capital landscape.

Kopelman explains his approach to early-stage investing and how First Round Capital has positioned itself in the competitive VC market. He discusses the importance of identifying promising startups with strong founding teams and disruptive business models.

The conversation delves into the challenges of scaling a venture fund while maintaining quality investments. Kopelman shares insights on portfolio construction, risk management, and the balance between supporting existing portfolio companies and finding new opportunities.

He also addresses the changing dynamics in the venture capital industry, including increased competition, the rise of micro VCs, and how technological advancements have impacted investment strategies and startup valuations.

Throughout the discussion, Kopelman emphasizes the importance of adding value beyond capital, building strong relationships with founders, and developing a distinctive investment thesis that guides decision-making in an increasingly crowded market.

Read More

A CEO of AI Applications Marks a New Era of AI Competition

Tomtunguz • which foundational model companies compete. • May 7, 2025

Technology•AI•AIApplications•Personalization•ModelCompetition•Essays


With models rapidly commoditizing in performance, we are seeing different strategies for keeping users on the same models.

MidJourney is asking users to personalize their models. OpenAI has developed a memory. Google is offering free fine-tuning as a way of winning share and locking in demand. OpenAI has hired Fidji Simo as CEO of Applications.

Ultimately, personalization and applications are likely to be the two vectors by which foundational model companies compete.

A model remembering your name, optimized to your codebase, having knowledge of your previous work and refashioning itself to the way you work: those are reasons for loyalty, even if the model isn’t state of the art. Like airplane reward programs, personalization and memory introduce switching costs that may outweigh the benefits of state-of-the-art models.

Fidji’s hiring and role suggest Applications for OpenAI are of paramount importance.

The title CEO of Applications implies Fidji will be the head of a substantial and strategic business unit and OpenAI will evolve to a holding company structure—like Sundar’s Alphabet with a CEO of Waymo (actually co-CEOs!) and YouTube (Neal Mohan) for example.

The personalization of the model within an application and workflow will be the ultimate moat. OpenAI’s recent leadership hire makes plain how large the opportunity is.

A recent paper on AI competition suggests that newer models are rapidly adopted by customers. Sometimes newer models sap share from competitors, other times replace previous versions (Claude 3.5 users upgraded to 3.7), last new models expand usage (Gemini Flash & 2.5).

Read More

Everyone Is Cheating Their Way Through College

Nymag • James D. Walsh • May 7, 2025

Education•AI•AcademicIntegrity•GenerativeAI•HigherEducation•Essays


Overview of AI Cheating in College

The article explores the widespread use of generative AI tools, especially ChatGPT and similar platforms, for cheating throughout higher education. It begins with the story of Chungin “Roy” Lee, a Columbia University computer science major who openly admits to using AI to complete 80% of his assignments without engaging deeply in the learning process. Lee’s experience reflects a larger trend: students across elite institutions, state schools, and community colleges are increasingly relying on AI to bypass traditional efforts in essays, coding, and exams. The article highlights how such AI tools are not only used to cheat on homework and papers but are also incorporated to pass technical interviews and even professional tests.

Student Perspectives and AI Dependency

Students like Lee and others, including Sarah from Wilfrid Laurier University and Wendy, a finance major, describe AI as indispensable to their academic success. For instance, Sarah notes her grades soared after initially using ChatGPT in high school, continuing through college. Wendy explains how she manipulates AI to draft papers that take a fraction of the time compared to writing manually, yet expresses nostalgia for more traditional writing. Many students do not see using AI as cheating but as a necessity, given the academic system’s rigor and pressures. The ease of AI-generated content is shifting perceptions of what constitutes legitimate work, with many students admitting both reliance on and addiction to these tools.

Faculty Challenges and Institutional Responses

Professors and universities face tremendous challenges in combating AI-enabled cheating. While some have attempted to design “AI-proof” assignments such as in-person exams or oral defenses, these measures are limited and often impractical at scale. AI detectors like Turnitin offer some support by estimating the likelihood of AI involvement in student submissions, but they frequently produce false positives or false negatives, especially affecting neurodivergent students and non-native English speakers. Faculty admit detecting AI-generated work is difficult, with many resorting to embedding obscure phrases in assignments to catch AI writers, though students quickly learn to circumvent these traps. The widespread nature of AI usage has caused frustration among educators, some of whom have become disillusioned enough to consider leaving academia.

Implications for Education and Skill Development

The article critically examines the long-term implications of AI dependency on student learning and skill acquisition. Experts warn of a potential generation emerging with superficial literacy and diminished critical-thinking skills. The substitution of human effort with AI can degrade memory, problem-solving, and creativity, accelerating a decline in intellectual development that predates AI’s rise. The transactional nature of college education—viewed more as a means to a high-paying job than intellectual growth—has been exposed by AI’s ability to complete academic tasks effortlessly. This calls into question the future readiness of graduates in the workforce and society if foundational skills like writing and analytical reasoning are compromised.

Evolving Definitions of Cheating and the Role of AI

The definition of cheating is becoming increasingly complex as AI becomes more integrated into education. OpenAI's CEO Sam Altman suggests that traditional notions of writing essays are outdated and that the educational system must evolve with technological changes. However, there remains concern about students disengaging from their own critical thinking and learning processes. The article shares diverse student viewpoints on AI usage’s ethical boundaries, illustrating the unclear divide between assistance and plagiarism. Tools like Cluely, developed by Lee and his co-founder, aim to further embed AI in daily tasks, including exams and interviews, heralding a future where AI may undercut traditional educational methods entirely.

Conclusion: A Crisis and a Call for Rethinking Education

The rise of generative AI in academia has plunged educators into an existential crisis, questioning the value, purpose, and structure of higher education. With rampant AI use accelerating academic dishonesty, educational institutions are confronting significant difficulties in maintaining integrity, assessing true student ability, and preparing students for meaningful futures. The article underscores that systemic reforms are necessary, including reconsidering how learning is assessed and valued. Until then, AI continues to "hack" education, blurring lines between human and artificial effort, and challenging personal, institutional, and societal notions of knowledge and achievement.

Read More

Venture Capital

Unshackled Founder Sentiment: Detailed Analysis of Confidence and Outlook

Unshackledvc • Manan Mehta • May 6, 2025

Business•Startups•ImmigrantFounders•FounderSentiment•FundraisingChallenges•Venture Capital

Unshackled Founder Sentiment: Detailed Analysis of Confidence and Outlook

In April 2025, Unshackled Ventures anonymously surveyed 32 immigrant startup founders across its portfolio to gauge their sentiments amidst evolving economic, immigration, and trade conditions. The findings highlight a community exhibiting strong short-term confidence in their ability to operate within the U.S., yet facing significant challenges in fundraising due to macroeconomic headwinds.

A majority of founders express strong confidence in their immediate ability to stay and build in the U.S., with 84% rating their short-term confidence as high (scores 4 or 5). Confidence remains generally positive over the long term—65% rate themselves as confident or very confident. However, 25% report neutral sentiment, and 9% express concern, pointing to growing uncertainty about what lies ahead, particularly for those navigating early-stage growth and temporary visa categories.

The predominant challenge identified is the difficulty in securing funding, surpassing concerns over immigration policies. Founders cite a tightened investment landscape and prolonged fundraising cycles as significant hurdles. Compared to previous periods, immigration-related anxieties have diminished, with many founders adapting to existing policies and focusing more on immediate business challenges. Despite obstacles, founders demonstrate a pragmatic approach, adjusting strategies to navigate the current economic climate and maintain business momentum.

The report presents a detailed analysis based on the Unshackled Founder Sentiment Survey conducted in April 2025. It engaged 32 founders across diverse sectors (Deeptech, Robotics, AI, AR/VR, Health and Digital Health, CPG, Agtech, Fintech, Cleantech, Hardware, Software/SAAS, Consumer, Enterprise, Ecommerce), company stages (from pre-seed to Series B+), and immigration pathways. Understanding the lived experience of these entrepreneurs—in all its complexity—is paramount. The analysis delves into their confidence levels, economic outlook, experiences with immigration policy, and overall mindset, aiming to provide unfiltered insights into how founders navigate today’s climate of uncertainty, innovation, and transformation.

Founders express notable confidence in their immediate ability to operate in the U.S., contrasting with greater uncertainty regarding their long-term prospects. Confidence is strongly skewed towards the positive end in the short term: 50% of founders report being very confident (score 5), and another 34% report being confident (score 4). Critically, no founders reported negative confidence (scores 1 or 2) regarding their short-term prospects, indicating widespread immediate operational stability.

In the long term, confidence levels show a broader spread: a majority still express high confidence (34% score 5, 31% score 4), but neutral confidence increases significantly to 25%, and 9% report negative confidence, highlighting underlying anxiety about the future. Variations exist by company stage, industry group, and immigration status, with certain groups like seed stage founders and student visa holders showing more tempered or neutral outlooks.

Read More

The Most Useful Type Of Feedback

Youtube • Uncapped with Jack Altman • April 18, 2025

Business•Management•Feedback•Leadership•Communication•Venture Capital


Watch this video on YouTube

Read More

Benchmark vs a16z: Why Stage Specific Firms Win

Youtube • 20VC with Harry Stebbings • May 8, 2025

Business•Startups•VentureCapital•Investment•Strategy•Venture Capital


In the venture capital world, stage-specific firms like Benchmark and a16z demonstrate distinct advantages compared to generalist or multi-stage firms. Benchmark focuses exclusively on early-stage investing, typically leading seed and Series A rounds. This concentrated approach allows them to build deep expertise in identifying and supporting startup founders at the critical inception phase. Their early-stage focus grants them access to the initial value creation in startups, helping shape company cultures and business models from the ground up.

On the other hand, a16z operates across multiple stages from seed to growth. This broad scope enables them to support companies throughout their lifecycle by providing a wide range of resources and expertise tailored to each growth phase. While this versatility is valuable for scaling companies, it can lead to less specialization at any single stage. Their strength lies in leveraging extensive operational resources and a robust network, which helps portfolio companies navigate growth challenges and market expansion.

Stage-specific firms like Benchmark tend to win by developing a reputation as experts in a particular phase, which attracts the most promising entrepreneurs early on. This early relationship-building creates a strong founder-firm bond and often results in significant financial returns when these companies succeed and scale. Conversely, multi-stage firms like a16z supplement early investments with follow-on rounds, offering continuity and support that can sustain companies through rapid growth and competitive environments.

In summary, the success of stage-focused venture firms stems from concentrated expertise and early engagement in startups, while multi-stage firms leverage comprehensive support and scalability services. Each model serves different purposes in the ecosystem but the specialized approach often leads to winning the best early-stage deals.

Read More

Bucky Moore @ Lightspeed Venture Partners: Why You Cannot Do VC If You Do Not Do Pre-Seed

Youtube • 20VC with Harry Stebbings • May 5, 2025

Business•Startups•VentureCapital•PreSeed•InvestmentStrategy•Venture Capital


Bucky Moore, a general partner at Lightspeed Venture Partners, emphasizes the critical importance of pre-seed investing in venture capital. He shares his perspective that if a VC firm is not active in the pre-seed stage, they are missing a foundational part of the venture ecosystem. Moore argues that early engagement with startups at the pre-seed stage allows a VC to build a deeper understanding of the company and its founders, creating a stronger basis for future investment rounds.

According to Moore, pre-seed investing is not only about securing early stakes in promising companies but also about cultivating long-term relationships and trust. He explains that this stage is often where the riskiest bets are made, but it’s also where the greatest value-add can come from the hands-on support and guidance that VCs provide. This involvement helps shape the product, hiring decisions, and overall strategy, often making a significant difference in the trajectory of the startup.

Moore highlights that successful pre-seed investing requires patience, a willingness to embrace uncertainty, and a focus on potential rather than immediate metrics. He sees it as a vital gateway into the venture capital world, where recognizing and backing the right teams early can lead to outsized returns.

The conversation also touches on how Lightspeed has structured its approach to pre-seed investing, combining deep founder empathy with rigorous market analysis. This enables them to back exceptional teams with innovative ideas that might not yet have fully proven concepts or traction.

By investing at the pre-seed stage, Lightspeed aims to build a pipeline of future leaders in the startup ecosystem, creating continuity and long-term value for both the fund and the entrepreneurs they support.

Read More

Fat vs Lean Startups

Youtube • Uncapped with Jack Altman • April 21, 2025

Business•Startups•Strategy•GrowthManagement•FundingModels•Venture Capital


In the startup world, two contrasting approaches often determine how companies allocate resources and manage growth: fat startups versus lean startups.

A fat startup typically raises a large amount of capital early on. This influx of funds allows the company to scale quickly, invest heavily in marketing, product development, and hiring. The strategy behind this approach is to capture market share rapidly and try to outpace competitors by leveraging substantial financial resources.

On the other hand, a lean startup takes a more cautious path. It focuses on minimizing expenses, conserving cash flow, and validating business ideas through iterative experimentation with minimal investment. This approach emphasizes testing market demand before scaling operations. Lean startups aim to become sustainable early on by making small, calculated bets and adjusting based on customer feedback.

The choice between fat and lean startups depends on market conditions, the nature of the product, and the founders' risk tolerance. Fat startups may face pressure to achieve high growth quickly due to investor expectations, which can lead to operational strain. Lean startups might grow more slowly, but they typically maintain tighter control over costs and have greater flexibility to pivot.

Both strategies have their merits and challenges. In some cases, a lean startup that proves its concept might seek additional funding later to scale aggressively, blending aspects of both models. Conversely, some fat startups may need to adopt lean principles later to optimize efficiency.

Ultimately, understanding these approaches helps founders and investors make informed decisions about resource allocation, growth strategies, and risk management.

Read More

Global Venture Funding Slowed In April, Despite Strong AI Showing

Gené Teare • Crunchbase • May 5, 2025

Business•VentureCapital•ArtificialIntelligence•Healthcare•InvestmentTrends•Venture Capital


Global venture funding in April 2025 reached $23 billion, maintaining the same level as the previous year but significantly decreasing from the $68 billion invested in March. This decline is attributed to macroeconomic challenges, including high interest rates and geopolitical tensions. Despite the overall downturn, the artificial intelligence (AI) sector demonstrated resilience, securing substantial investments.

AI companies attracted approximately $5.7 billion in funding, accounting for about 25% of the total venture capital invested globally. This marks a notable increase from previous months, highlighting the growing confidence in AI's potential. The healthcare sector also saw significant investment, with companies raising around $5.7 billion, representing 25% of the total funding. These sectors have been pivotal in sustaining venture capital activity amid broader economic uncertainties.

The funding landscape was characterized by a concentration of large deals. Notably, 21 companies secured funding rounds valued at $1 billion or more, indicating a continued appetite for substantial investments in high-growth potential startups. This trend underscores the market's focus on scaling companies capable of delivering significant returns.

In contrast, other sectors experienced a decline in investment. The fintech industry, for instance, saw a reduction in funding compared to the same period in the previous year, reflecting a cautious approach from investors in certain areas. This shift suggests a reevaluation of risk and a preference for sectors perceived as more resilient or promising.

Overall, while global venture funding in April 2025 faced challenges, the strong performance of AI and healthcare sectors highlights the dynamic nature of the investment landscape. The market continues to adapt, with investors gravitating towards industries that offer innovation and growth opportunities.

Read More

Running Y Combinator Like a Founder | Garry Tan

Youtube • Uncapped with Jack Altman • April 17, 2025

Business•Startups•Entrepreneurship•Accelerators•Mentorship•Venture Capital


Garry Tan, a founder and investor known for his work with Y Combinator and Initialized Capital, shares insights on running Y Combinator with the mindset of a founder. He emphasizes the importance of empathy and understanding the challenges startup founders face in order to better support and guide them through the accelerator program.

According to Tan, running an accelerator like Y Combinator requires a deep commitment to the founders' success. It’s about creating an environment that fosters innovation, rapid iteration, and resilience. This approach means not just providing capital and resources, but also mentoring, advice, and a community that founders can rely on.

Tan highlights the iterative nature of startups, where constant learning and adapting are critical. He points out that founders often need to make tough decisions quickly, pivot ideas based on feedback, and persevere through failures. YC's role is to facilitate this process by sharing knowledge, offering practical support, and encouraging experimentation.

The discussion also touches on the importance of building a strong, diverse community of startups within the program. This network effect allows founders to learn from each other, find collaborators, and create a sense of belonging that fuels motivation and innovation.

By treating Y Combinator like a startup itself, Garry Tan explains that the leadership team can continually improve the program, experiment with new initiatives, and stay closely connected with the changing landscape of entrepreneurship and technology innovation.

He also reflects on the balance needed between structure and flexibility in the accelerator experience, enabling founders to explore their vision while providing enough guidance to keep them on a productive path.

Read More

Investing In Themes Means More

Youtube • Uncapped with Jack Altman • April 14, 2025

Finance•Investment•ThematicInvesting•GrowthStrategies•PortfolioManagement•Venture Capital


Investing in thematic trends provides a powerful way to capture broad, structural growth that transcends individual companies. The approach focuses on long-term megatrends shaping the future economy rather than trying to pick the winners in isolated sectors or short-term bets.

Themes often encapsulate multiple sectors and industries impacted by a common driver like technological innovation, demographic shifts, or regulatory changes. This holistic lens can identify durable growth patterns, reduce dependency on single companies, and tap into evolving consumer or business behaviors globally.

For example, themes around remote work and digital collaboration gained heightened importance over recent years as broader societal and technological changes accelerated adoption. Investing with a thematic mindset involves a mix of strategic foresight and the discipline to navigate volatility tied to evolving narratives.

Investors embracing thematic strategies consider how integrations of AI, sustainability, health tech, or automation will unfold over years or decades. They build exposure across related businesses positioned to capture different parts of these ecosystems, creating diversified portfolios aligned with future market transformations.

In essence, thematic investing is about more than just finding the next hot stock; it’s about identifying meaningful shifts shaping investment opportunities at scale and staying engaged through their evolution to realize substantial growth.

Read More

The Power of Transferring Belief

David Cummings • May 3, 2025

Business•Entrepreneurship•Leadership•Communication•Pitching•Venture Capital


Last week, I listened to an entrepreneur deliver his pitch, and by the end, I was struck by how deeply he believed in his company's vision. His conviction was contagious, transforming his belief into mine. All entrepreneurs believe in their work, but some have a unique ability to persuade others to share their perspective.

I call this "transferring belief" — the capacity to convey your vision so convincingly that others adopt it as their own. It's a fundamental skill for entrepreneurs, especially when pitching to investors, recruiting team members, or selling to customers.

The most successful entrepreneurs I've met possess this ability to transfer belief. They articulate their vision with such clarity and passion that it becomes irresistible. This isn't just about charisma; it's about authentic conviction combined with the ability to communicate effectively.

When an entrepreneur truly transfers belief, they create evangelists who champion their cause, investors who back their vision, and customers who become loyal advocates. This transferred belief becomes a powerful force multiplier for the business.

To develop this skill, entrepreneurs should focus on deeply understanding their "why," crafting a compelling narrative, and consistently reinforcing their vision through actions and communications. The most effective belief transfers happen when the entrepreneur's conviction is genuine and their vision addresses a meaningful problem.

Read More

Geopolitics

The Greed & Fear Tango: The Markets in April 2025!

Blogspot • Aswath Damodaran • May 3, 2025

Finance•Investment•MarketVolatility•Tariffs•GlobalEconomy•Geopolitics


I started the month on a trip to Latin America, just as the tariff story hit my newsfeed and the market reacted with a sell off that knocked more than $9 trillion in market cap for global equities in the next two days. The month was off to a bad start, and tariffs remained the lead story for much of the month, contributing to both its biggest down days (with stories of trade war escalation) and to the biggest up days (with news of relief from the fight). To add to the volatility, there was talk midway through the month of replacing Jerome Powell as the Fed Chair, and assorted news adding to uncertainty about the direction of the economy. An observer reading just the news stories and asked to guess what the market did during the month would probably have bet on stocks falling steeply, but he or she would have lost that bet, because markets managed to surprise us all again, ending the month almost where they began.

Equities: Storm Clouds gather (and dissipate)!

It would be an understatement to describe equity markets in April 2025 as volatile, with the equity indices going through stomach wrenching up and down movements intraday and across days, as investors struggled to price in a world of tariffs, trade wars and policy uncertainty.

The journey that the indices went through during the course of the month has been extraordinary. Each of the indices lost close to 10% in the first two days of the month, went deeper into the hole in the second week of the month, but by the end of the month, they had each found their way back to almost where they started the month at, with the S&P 500, NASDAQ and the MSCI world index all within 1% of their start-of-the-month levels.

As I noted in my post right after the first couple of days of this month, where I framed the market crisis around tariffs, the indices sometimes obscure markets shifts that are occurring under the surface, and that looking at all publicly traded stocks on an aggregated basis provides a more complete picture. I will start by looking at the regional breakdown of what the month delivered in terms of change in market cap, in both dollar and percentage terms:

The crisis may have been birthed in the United States, but as has been the case with market crises in this century, it has spread across the world, with disparate impacts. There are truly no standouts in either direction, with China being the worst performing region, in terms of percentage change in dollar value, down 3.69%, and India and Latin America tied for best performing, up 3.57%. These are dollar returns, and since the US dollar came under selling pressure during the course of the month, the local currency returns were worse, especially in markets, like the EU, where the Euro gained about 5% in the courser of the month.

Read More

Bernstein's Stacy Rasgon: If Nvidia can't play in China we'll hand the Chinese AI market to Huawei

Youtube • CNBC Television • May 6, 2025

Technology•AI•Semiconductors•ChinaMarket•Huawei•Geopolitics


If Nvidia is restricted from operating in China, the Chinese AI market is likely to be dominated by local companies such as Huawei, according to Stacy Rasgon, an analyst at Bernstein. Rasgon highlights the importance of Nvidia's role in AI chip supply worldwide, and a potential limitation on Nvidia’s access to China could shift the regional competitive landscape significantly.

Nvidia's GPUs are foundational to major AI developments, powering many deep learning models, and thus, any restrictions imposed on their sale in China would create a substantial vacuum. This vacuum would be an opportunity for Chinese companies, notably Huawei, to increase their market share by stepping in as alternative suppliers of AI hardware.

The Chinese government has been heavily investing in developing its domestic chip manufacturing capabilities, aiming to reduce reliance on foreign technology amid rising geopolitical tensions. This increasing emphasis on self-sufficiency in semiconductor technology is further amplified by the increasing global competition in AI infrastructure.

Rasgon explains that Nvidia's absence in the Chinese market would likely accelerate Huawei’s rise, as Huawei already has strong government backing and a solid foundation in telecommunications and chip design. This scenario highlights the broader implications of tech decoupling between the U.S. and China, especially in the critical and fast-evolving AI sector.

Such shifts could have profound consequences not only for market competition but also for the pace and direction of AI innovation globally, as fragmentation in access to critical technology components becomes more pronounced.

Government Overreach

US to demand Google break up its online advertising business

Ft • May 2, 2025

Technology•Software•Google•Antitrust•DigitalAdvertising•Government Overreach


The US Department of Justice is seeking to break up Google's advertising technology business following its recent trial victory against the tech giant. The department has asked a federal judge to force Alphabet, Google's parent company, to sell off key components of its online advertising infrastructure, including its ad exchange and publisher server technology.

This move represents one of the most significant antitrust actions against a major technology company in recent years, with the potential to fundamentally reshape the digital advertising landscape. The Justice Department's request follows its successful case arguing that Google has monopolized the online advertising market.

The components targeted in the breakup request include Google's ad exchange, which functions as a marketplace where advertising space is bought and sold in real-time auctions, and its publisher server technology, which websites use to manage and sell their advertising inventory.

If approved by the court, this forced divestiture would dramatically alter Google's business model, which has relied heavily on its integrated advertising technology stack to generate billions in revenue. Critics of Google have long argued that the company's control over multiple layers of the advertising technology ecosystem creates inherent conflicts of interest and allows it to favor its own products.

The case represents part of a broader push by US regulators to address perceived monopolistic practices by major technology companies. Google has previously indicated it would appeal any adverse ruling in the case.

Read More

AI

OpenAI Attempts to Have Their For-Profit Cake and Eat It Too

Spyglass • M.G. Siegler • May 5, 2025

Technology•AI•OpenAI•CorporateStructure•Governance


It's not exactly shocking, but it is surprising. OpenAI has effectively laid down their arms in the fight to become a for-profit. Instead, they'll seek to move to a structure they believe will get most of what they wanted out of such a maneuver, without fully maneuvering it.

At least, that's how it sounds right now. I suspect there's going to be a lot of back and forth around all of this in the coming days, weeks, and months. But clearly Sam Altman believes – or wants everyone else to believe – that shifting OpenAI's for-profit arm to a Public Benefit Corporation, while still allowing the non-profit to have effective control over that entity, will square the circle. We'll see.

What's clear is that OpenAI thought they were not going to win the battle to invert the company and become a full for-profit with a non-profit arm. Elon Musk threw a wrench in this plan, but many others held wrenches here too. As it turns out, there's a lot at stake and a lot of potential stakeholders when you're talking about a $300B non-profit!

Removing the cap on the potential profits that the non-profit would potentially earn one day (lol) seems straightforward enough as that whole structure was silly to begin with. But less straightforward is just about everything else here, most of which isn't stated in OpenAI's blog post on the matter. Will those holding the current rights to future profits be converted into actual equity holders in some way? This article by Cade Metz of The New York Times suggests that may happen:

"OpenAI said that it was still negotiating the nonprofit's stake in the new corporation and that the nonprofit would pick the board members of the new entity."

It implies that the non-profit would hold an equity stake in the new PBC, and that such a stake would give it control. But there's a lot of variables in even just that. Would this non-profit hold 51% of the voting rights? More? Is that the same as the equity structure or would the non-profit be granted some kind of shares with out-sized voting rights? Or if it's straight equity control, might they be diluted below a controlling stake in a future fundraise?

Where does this leave Microsoft? The official blog post cites them once: "We made the decision for the nonprofit to stay in control after hearing from civic leaders and having discussions with the offices of the Attorneys General of California and Delaware. We look forward to advancing the details of this plan in continued conversation with them, Microsoft, and our newly appointed nonprofit commissioners."

Reading this, I certainly don't get the sense that OpenAI itself is fully confident in all of this – hence the seemingly out-of-the-blue blog post on the matter backtracking from the previously confident stance on the matter.

The first line of Altman's blog post is the only one I can cite with confidence: "OpenAI is not a normal company and never will be."

Read More

What Is an AI Agent?

Youtube • a16z • May 2, 2025

Technology•AI•MachineLearning•SoftwareDevelopment•Innovation


In this episode of AI + a16z, a16z Infra partners Guido Appenzeller, Matt Bornstein, and Yoko Li discuss and debate one of the tech industry's buzziest words right now: AI agents. The trio digs into the topic from multiple angles, including whether a uniform definition of "agent" exists, how to distinguish agents from LLMs and software functions, and the practical challenges of pricing and deploying agent systems.

They explore the continuum of agent capabilities, from simple prompt-driven interfaces that mimic human interactions to more sophisticated systems capable of autonomous task execution. The discussion highlights disagreements in defining agents, with some interpretations including basic knowledge-base chatbots as agents if they present a human-like interaction model.

Key questions addressed include whether agents can truly replace humans in workflows, how to differentiate agentic behavior from traditional software functions, and the implications of web access for agents operating in data-siloed environments. The conversation also examines pricing models, emphasizing the need to align costs with measurable value rather than technical complexity.

Read More

The Deadly “AI Slow Roll” in SaaS: It May Cost You Everything

Saastr • Jason Lemkin • May 5, 2025

Technology•AI•SaaS•BusinessStrategy•Productivity


I'm seeing an alarming trend across the SaaS landscape that's going to separate winners from losers in the next 18 months: The AI Slow Roll.

You know what I'm talking about:

  • It's the VP of Engineering who's "still evaluating Windsurf and Cursor" but says it's still a little bit early.

  • It's the Product leader confidently declaring AI "isn't ready for prime time" because of hallucinations.

  • It's the CRO insisting that "customers all want to talk to a human" and that "a great sales rep beats AI any day of the week."

Let me be crystal clear: Change fast or die. At least, watch your business atrophy.

Everyone that's invested in a wide group of B2B and SaaS companies is seeing this. For the most part, those that haven't aggressively adopted AI in their products are seeing growth slow. Sometimes, they are seeing a bit uptick in churn. Sometimes, retention is staying up but they are losing deals to new entrants. Whatever it is, it's a threat if you aren't ahead of the curve here. And it's accelerating.

Is your team really doing what you need it to here? If not, you will lose.

This may sound dramatic, and honestly I wasn't even sure change would come this fast just a few months ago. But now it's 100% clear.

The Best Dev Teams Are Already Automating 30%-50% of Code Into Production. And Shipping So Many More Features

While many are still "evaluating options," your competitors have already transformed their operations:

  • Top-tier tech companies are automating 50% of their code production with AI tools

  • Even conservative enterprises like Salesforce have hit 20% automation

  • HubSpot is producing so many new features it can't even put them all into production anymore

  • The gap between AI-native and AI-hesitant companies is widening every single week

Companies deploying AI at scale are seeing 30-40% productivity gains in engineering and customer satisfaction scores increasing when AI is properly deployed in support. These are the two areas furthest ahead in AI. But sales, marketing and success are not far behind.

"Hallucinations" Are Just an Excuse for Inaction

When executives tell me "we're concerned about hallucinations," what I really hear is: "we don't want to do the work."

This problem is already solved by companies who approach AI deployment seriously:

  • Implement proper training and fine-tuning

  • Build exception handling into your workflows

  • Create human escalation paths for edge cases

  • Start with non-critical use cases and expand

Read More

What OpenAI’s restructuring plan means for its corporate future

Techcrunch • Maxwell Zeff • May 6, 2025

Technology•AI•OpenAI•Restructuring•PublicBenefitCorporation


OpenAI has announced a restructuring plan to transition its for-profit arm into a Public Benefit Corporation (PBC), while maintaining control under its nonprofit board. This decision follows discussions with attorneys general from Delaware and California, who had been closely monitoring OpenAI's corporate structure. The new plan aims to balance the company's rapid growth and the need for additional capital with its mission to develop artificial general intelligence (AGI) safely and for the benefit of humanity.

The restructuring involves converting OpenAI's for-profit subsidiary into a PBC, a hybrid entity that allows for-profit operations while upholding a public mission. The nonprofit board will continue to oversee the PBC, ensuring that OpenAI's activities align with its foundational goals. This move is intended to attract further investment, potentially from firms like SoftBank, while maintaining accountability to the public benefit rather than solely to shareholders.

This decision comes after legal and ethical concerns, including a lawsuit from co-founder Elon Musk and scrutiny from state attorneys general. Critics have questioned whether the new structure sufficiently protects OpenAI's charitable objectives. Despite such concerns, CEO Sam Altman emphasized that the hybrid structure is key for securing further investment, potentially from firms like SoftBank, while ensuring accountability to public benefit rather than just shareholders.

The shift seeks to align OpenAI’s rapid growth—highlighted by a $300 billion market valuation and 400 million weekly ChatGPT users—with its founding mission to advance artificial general intelligence (AGI) safely. The company had faced legal and ethical concerns, including a lawsuit from co-founder Elon Musk and scrutiny from state attorneys general. Critics have questioned whether the new structure sufficiently protects OpenAI's charitable objectives. Despite such concerns, Altman emphasized that the hybrid structure is key for securing further investment, potentially from firms like SoftBank, while ensuring accountability to public benefit rather than just shareholders.

Read More

Nvidia CEO Jensen Huang: It would be a tremendous loss not to be able to address China's AI market

Youtube • CNBC Television • May 6, 2025

Technology•Hardware•AI•GeopoliticalTrade•Semiconductors


In this interview, Nvidia CEO Jensen Huang discusses the impact of US restrictions on China's AI market. When asked about the possibility of further restrictions on chip sales to China, Huang emphasizes the significance of the Chinese market.

"It would be a tremendous loss not to be able to address the China market," Huang states. He explains that China represents roughly 20-25% of the world's GDP and is home to many important companies and academic institutions.

Huang acknowledges the legitimate concerns of the US government regarding national security but expresses hope that the industry can find a way to address these concerns while still maintaining access to the Chinese market. He notes that Nvidia has been working closely with the US government to ensure compliance with regulations.

The CEO also discusses the broader AI landscape, highlighting Nvidia's role in enabling the AI revolution through its advanced GPU technology. He explains how Nvidia's hardware and software ecosystem has become essential infrastructure for AI development worldwide.

When asked about competition, Huang maintains confidence in Nvidia's position but acknowledges that the company must continue to innovate to stay ahead. He discusses how Nvidia's integrated approach to AI computing gives it an advantage over competitors focusing on individual components.

Read More

Google shares drop after Apple says default Safari searches dipped for first time in April

Youtube • CNBC Television • May 7, 2025

Technology•Web•SearchEngines•Google•Apple•AI


Shares of Alphabet, Google's parent company, dropped after Apple reported that the number of default Safari searches conducted with Google declined for the first time in April. This shift signals a potential increase in competition for Google's crucial search business, as Apple has been diversifying its search engine partners.

Apple's Safari browser, the default on all Apple devices, traditionally defaults to Google for search queries. However, recent data shows a dip in Google's default search volume on Safari, which may reflect growing user adoption of alternative search engines or changes in default settings. This development could impact Google's ad revenue, given the significance of search traffic from Apple devices.

Industry analysts are watching this closely as it marks a rare disruption in Google's dominance of search traffic from Apple users. The interplay between Apple and Google's search preferences highlights the evolving dynamics in the tech industry, where platform owners like Apple shape user behavior by selecting specific default services.

Alphabet's shares decline following this announcement suggests investor concern about the potential revenue impact. While Google remains the predominant search engine worldwide, changes in default search engine use represent a strategic challenge for the company moving forward.

Read More

Alphabet shares plunge as Apple seeks AI alternatives to Google search

Ft • May 7, 2025

Technology•AI•Search•Apple•Google


Apple is exploring alternatives to Google's search engine for its iPhone and Safari browser, including AI-powered options from start-ups, according to a senior company executive. This revelation has sent Alphabet shares plunging as it raises concerns about the future of the lucrative partnership between the two tech giants.

Google currently pays Apple billions annually to be the default search engine on Apple devices. This arrangement has been a significant revenue stream for Apple and a crucial source of traffic for Google's search business. However, with the rise of AI technology, Apple appears to be reconsidering its options.

The senior Apple executive indicated that the company is evaluating various AI-powered search alternatives from technology start-ups. This move comes as part of Apple's broader strategy to enhance its AI capabilities and potentially reduce its dependency on Google.

Industry analysts suggest that Apple's interest in alternative search providers could be motivated by several factors, including the desire to offer more privacy-focused search options to its users, to develop a competitive edge in AI services, and to diversify its revenue streams beyond its current partnership with Google.

The news has had an immediate impact on the market, with Alphabet's stock experiencing a significant drop. Investors are concerned about the potential loss of the Apple partnership, which drives substantial search traffic to Google and contributes significantly to its advertising revenue.

Read More

Microsoft adopts Google’s standard for linking up AI agents

Techcrunch • Kyle Wiggers • May 7, 2025

Technology•AI•OpenProtocol•AIAgents•Collaboration


Microsoft says that it’s embracing Google’s recently launched open protocol for allowing AI “agents” to communicate with each other. On Wednesday, Microsoft announced that it would bring support for Google’s Agent2Agent (A2A) spec to two of its AI development platforms, Azure AI Foundry and Copilot Studio. Microsoft has also joined the A2A working group on developing the protocol further.

The A2A specification is designed to enable disparate AI systems, or “agents,” to interoperate, share information, and collaborate efficiently across different platforms and environments. By adopting this open standard, Microsoft aims to enhance how its AI tools interact within its own ecosystem and with others outside it, creating a more integrated and flexible AI infrastructure.

The move signals a growing industry trend toward establishing universal standards for AI communication, enabling more complex and capable AI workflows that can leverage multiple AI models and services collaboratively. Microsoft's support for the A2A protocol not only facilitates cross-company AI collaboration but also underlines its commitment to open AI development and standardization.

This adoption is expected to accelerate the development of AI agents capable of working together in real-time, solving problems, and delivering richer outcomes through shared intelligence. Microsoft’s integration into the A2A working group also means it will have an active role in shaping the future evolution of this communication standard.

Read More

Modernizing Warfare and Beyond

Youtube • a16z • May 6, 2025

Technology•AI•DefenseTechnology•AutonomousSystems•CyberWarfare


Modern warfare is undergoing a transformation driven by advances in technology, reshaping the nature of conflict and security on a global scale. Emerging technologies such as artificial intelligence, autonomous systems, cyber capabilities, and advanced sensing are enabling new forms of combat and defense strategies that prioritize speed, precision, and adaptability.

At the core of this transformation is the integration of AI and machine learning to process vast amounts of data in real time, supporting decision-making and enabling predictive analytics for battlefield awareness. Autonomous platforms, including drones and unmanned vehicles, are extending the reach and persistence of military operations while reducing risk to human personnel.

In addition, cyber warfare has become a critical domain, where offensive and defensive operations target information infrastructure and digital assets, impacting both military and civilian systems. This shift demands new doctrines and approaches to maintain superiority in an increasingly contested and complex environment.

The modernization of warfare also emphasizes interoperability and network-centric operations, where different systems and forces operate cohesively through secure communications and shared situational awareness. This enhances coordination and effectiveness across multi-domain battlefields, spanning land, air, sea, space, and cyberspace.

Beyond the direct application of these technologies, there is an ongoing evolution in how defense sectors approach innovation, collaboration, and procurement. Agile development processes and partnerships with commercial tech firms are accelerating the deployment of cutting-edge capabilities to meet rapidly changing threats.

Together, these developments signal a profound change not only in military operations but also in global security dynamics, requiring policymakers, technologists, and strategists to rethink traditional paradigms of deterrence, defense, and conflict resolution.

Read More

Anthropic launches Claude web search API, betting on the future of post-Google information access

Venturebeat • Michael Nuñez • May 7, 2025

Technology•AI•Web•SearchAPI•InformationAccess


Credit: VentureBeat made with Midjourney

Anthropic has launched a web search API for its AI assistant Claude, positioning itself to play a crucial role in the emerging landscape of post-Google information retrieval. The introduction of this API aligns with Apple's exploration of AI-powered search alternatives, signaling a broader industry shift in how users access and discover information online.

The Claude web search API enables developers to integrate contemporary, web-based search capabilities directly into AI interactions. This development moves beyond traditional static knowledge bases, allowing AI to access and synthesize real-time data from the internet, thus enhancing accuracy and relevance.

This launch comes at a pivotal moment as technology companies reconsider their reliance on Google for search, especially with the growing prominence of AI-driven tools. Anthropic’s offering exemplifies how AI can transform search dynamics by blending natural language understanding with live web data, potentially reshaping search engines and digital information ecosystems.

By tapping into the API, applications can deliver more mindful, context-aware responses that better reflect current events and diverse sources, ultimately offering users a more powerful and nuanced search experience. The move also highlights the competitive race to redefine search, involving major players seeking to innovate beyond the conventional search engine model.

Anthropic’s new web search API for Claude represents a fundamental step in advancing AI’s role in how digital information is accessed and delivered, marking notable progress in the shift toward AI-centric information discovery tools.

Read More

Claude Code: Anthropic's Agent in Your Terminal

Latent • to tell all! • May 7, 2025

Technology•AI•Software•DeveloperTools•Productivity

Claude Code: Anthropic's Agent in Your Terminal

The AI coding wars have now split across four battlegrounds: AI IDEs, vibe coding platforms, teammate agents, and CLI-based agents. CLI-based agents, such as OpenAI Codex and Claude Code from Anthropic, are notable for being composable and pay-as-you-go, with users charged based on tokens consumed.

Claude Code is described as less of a product and more like a Unix utility, following Anthropic's product principle to "do the simple thing first." This philosophy manifests in features like memory implementation through markdown files and simple prompt summarization approaches. It prioritizes minimal, understandable, and extensible building blocks, resembling the original UNIX design concept.

Unlike other products, Claude Code offers the most direct way to interact with Sonnet for coding, without hidden prompt optimization layers. This transparency is reflected in its average user spend of about $6 per day, much higher than competitors like Cursor. Inside Anthropic, some engineers have even spent over $1,000 in a single day, underscoring the tool's power and utility.

Claude Code runs in the terminal, giving it access to run bash commands and view files directly, enabling an agentic experience in a natural coding environment. Its development was somewhat experimental, starting with a version used for varied tasks before evolving into a coding tool widely adopted internally at Anthropic, signaling product-market fit.

The product team emphasizes simplicity and scrappiness, iterating quickly to see whether users truly engage before scaling. The tool is intended as an early, raw-stage product designed for power users who want direct model access rather than a polished UI experience.

Features developed since launch include web fetch, autocomplete, autocompact (for infinite context), auto-accept mode, and vim mode integrations. Claude Code itself writes most of its own code, with about 80-90% of the codebase being generated by the tool, aided by human reviews for complex areas.

For advanced use cases, Claude Code supports custom slash commands and integration with Anthropic’s MCP protocol, allowing versatile workflows without over-reliance on a single technology. It can integrate with GitHub actions, run semantic linting beyond traditional rules, and operate in both interactive and non-interactive modes. In non-interactive mode, users can automate tasks such as updating tests or generating changelogs efficiently.

The team also discusses the importance of permissions and safety controls, allowing fine-grained management of actions like file editing or running tests. Balancing autonomy and human oversight is central, especially given the non-deterministic nature of generative models.

Memory management follows the principle of simplicity, using markdown files automatically loaded to maintain context. Anthropic's approach to memory and context involves agentic search using existing code search tools rather than relying on separate indexed stores, reducing complexity and security risks.

Looking ahead, Claude Code aims to support varied workflows with sandboxing, branching, and rewind capabilities to improve development flexibility. The tool already supports parallel workflows where multiple agents explore different solutions, and plans to enhance planning and thinking tools rooted in chain-of-thought prompting.

Anthropic views Claude Code not just as a CLI tool but as part of a broader ecosystem and product suite optimized for developer productivity. The team is expanding and encourages anyone passionate about coding, LLMs, and terminals to connect and contribute to the project.

Read More

Eddy Cue, on the Stand in U.S. v. Google Trial, Says Apple Is Eying a Move to AI in Safari

Bloomberg • John Gruber • May 7, 2025

Technology•AI•Web•SearchEngines•BrowserInnovation


Apple Inc. is “actively looking at” revamping the Safari web browser on its devices to focus on AI-powered search engines, a seismic shift for the industry hastened by the potential end of a longtime partnership with Google.

Eddy Cue, Apple’s senior vice president of services, made the disclosure Wednesday during his testimony in the US Justice Department’s lawsuit against Alphabet Inc. The heart of the dispute is the two companies’ estimated $20 billion-a-year deal that makes Google the default offering for queries in Apple’s browser. The case could force the tech giants to unwind the pact, upending how the iPhone and other devices have long operated.

Cue said he believes that AI search providers, including OpenAI, Perplexity AI Inc. and Anthropic PBC, will eventually replace standard search engines like Alphabet’s Google. He said he believes Apple will bring those options to Safari in the future. “We will add them to the list — they probably won’t be the default,” he said, indicating that they still need to improve. Cue specifically said the company has had some discussions with Perplexity.

“Prior to AI, my feeling around this was, none of the others were valid choices,” Cue said. “I think today there is much greater potential because there are new entrants attacking the problem in a different way.”

If they can pay, Apple will listen. And I don’t think it’s bullshit, at all, that traditional web search is actually going into decline now because of AI. Honestly at this point it would be weird if it weren’t.

But. Let’s say Apple would prefer for the current arrangement between Apple and Google to continue as is. But it’s under threat as a remedy in Google’s monopoly case. Is this not the perfect testimony? Traditional web search is in decline, usage-wise — and Apple is considering deals with multiple upstarts. I think it’s all true. But I also think it helps make the case that the current deal between Apple and Google should not be disallowed.

I don’t think there’s any bullshit here. I think we’re at a highly competitive moment between browsers and old-school search and new-school AI. And I think Eddy Cue is right in the middle of it.

Read More

Good Vibrations Between Apple & Anthropic

Spyglass • M.G. Siegler • May 5, 2025

Technology•AI•Software•Programming•Apple


Well, one big internal change around Apple and AI didn’t take long...

Apple is teaming up with startup Anthropic on a new “vibe-coding” software platform that will use artificial intelligence to write, edit and test code on behalf of programmers.

The system is a new version of Xcode, Apple’s programming software, that will integrate Anthropic’s Claude Sonnet model, according to people with knowledge of the matter. Apple will roll out the software internally and hasn’t yet decided whether to launch it publicly, said the people, who asked not to be identified because the initiative hasn’t been announced.

This seems like exactly the type of thing they might want to announce at a certain developer-focused conference in June... Then again, we’re only weeks away and presumably Apple won’t have enough time to fully integrate the functionality by then (hence why it’s perhaps internal-only, for now). Do they dare risk pre-announcing something again? After all:

Last year, Apple announced its own AI-powered coding tool for Xcode called Swift Assist. The company had intended to roll it out in 2024 but never actually shipped it to developers. Internally, engineers have complained that the company’s own system could hallucinate — or make up information — and even slow down app development. The Anthropic partnership is an acknowledgment that Apple could use some outside help, though the two systems could ultimately work together.

Then again, Claude is fairly well tried and tested at this point. And many consider it to be the best model for "vibe coding". And this is a particular problem for Apple right now amongst their developer set. Riffing on this piece by Bryan Irace, I wrote the following back in March:

This feels especially timely given that the dates for WWDC – Apple's developer conference – were just announced. Apple has already taken the hit for postponing at least some of those user-facing AI features until later this year or next (or later), so it will put extra pressure on them to showcase to developers what their AI can do for them. They now have just over two months to polish that narrative – and they have to be extra careful here, because they really can't afford to show things that aren't ready to ship – in the fall, at the latest.

Beyond the consumer AI vaporware, Apple also somehow managed to bungle the roll-out of Swift Assist, shown off at last year's WWDC and still basically nowhere to be seen in the wild. Not great, Tim.

Again, this might give them a relatively straightforward way to rectify the situation. And it continues Apple down a rather interesting path of partnering on AI, rather than trying to go it alone. First, of course, came ChatGPT. And now, it seems, Gemini is closer to launching as well as a part of Apple Intelligence (we'll almost for sure hear more about it at WWDC). An Anthropic partnership on coding makes a lot of sense on the surface, the question is if Claude would also be a part of the more general "world knowledge" Apple Intelligence features – presumably Anthropic would want such a deal alongside the use of Claude for coding?

Then the question shifts back to my old favorite one: what kind of deals might Apple strike for these features and functionality? The OpenAI partnership nearly shifted from free (as in beer) to paid for via an investment before that fell apart. But now there's seemingly some sort of rev share agreement in place? For Google, all you can be sure about is that whatever the deal is, it will not be exclusive! For Anthropic, might there be some sort of investment angle there? Or just straight cash, homey?

Read More

DeepSeek. Temu. TikTok. China Tech Is Starting to Pull Ahead.

Nytimes • May 4, 2025

Technology•AI•ChinaTech•Innovation•DigitalEconomy


America must discard the belief that it is beating China in the innovation race.

While U.S. tech companies have been making headlines with their advancements in AI and other technologies, China has been quietly building formidable competitors across multiple domains. DeepSeek, a Chinese AI startup founded by former ByteDance engineers, recently released a large language model that rivals those from OpenAI and Anthropic in certain benchmarks, demonstrating China's growing capabilities in artificial intelligence research.

Meanwhile, Temu has emerged as one of the fastest-growing e-commerce platforms globally, challenging established players like Amazon with its ultra-low prices and direct-from-manufacturer business model. The app reached 100 million U.S. downloads faster than any other shopping app in history, showcasing China's prowess in consumer technology and supply chain innovation.

TikTok continues to dominate social media despite ongoing regulatory challenges in the United States, with its recommendation algorithms and engagement metrics outperforming Western competitors. The platform's parent company, ByteDance, has leveraged its AI expertise to create an entertainment ecosystem that captures unprecedented user attention and data.

These examples illustrate a concerning trend: China's technological capabilities are advancing rapidly across multiple sectors, from fundamental AI research to consumer applications. The notion that U.S. technical superiority is assured or that Chinese innovation is merely imitative no longer holds true in many critical domains.

American policymakers and business leaders must acknowledge this reality and respond with increased investment in research, education, and strategic industrial policy rather than complacency based on outdated assumptions about the global technology landscape.

Read More

OpenAI ditches plan to convert to for-profit business

Ft • Elon Musk • May 5, 2025

Technology•AI•OpenAI•Nonprofit•PublicBenefitCorporation


OpenAI has decided to maintain its nonprofit governance structure, reversing a prior plan to convert into a for-profit entity. This decision follows significant criticism, notably from co-founder Elon Musk and members of the AI academic and former employee communities. The proposed restructuring had raised concerns that for-profit status would compromise OpenAI’s founding mission to ensure AI benefits humanity. (ft.com)

Instead, OpenAI will convert its for-profit subsidiary into a public benefit corporation (PBC), allowing more flexible investor returns while maintaining control with the nonprofit. CEO Sam Altman emphasized that the decision was not driven by external pressure. The move also comes after complex discussions with state attorneys-general from Delaware and California, aiming to ensure OpenAI remains committed to its charitable goals. (apnews.com)

Board Chair Bret Taylor confirmed the final decision was shaped by feedback from civic leaders and legal dialogues. This resolution enables OpenAI to continue aligning its commercial structure with its foundational ethical mission under nonprofit oversight. (reuters.com)

Read More

Interview of the Week

Keen On America
Episode 2526: Keach Hagey on why OpenAI is the parable of our hallucinatory times
Much has been made of the hallucinatory qualities of OpenAI’s ChatGPT product. But as the Wall Street Journal’s resident authority on OpenAI, Keach Hagey notes, perhaps the most hallucinatory feature the $300 billion start-up co-founded by the deadly duo of Sam Altman and Elon Musk is its attempt to be simultaneously a for-profit and non-profit company…
Listen now

Much has been made of the hallucinatory qualities of OpenAI’s ChatGPT product. But as the Wall Street Journal’s resident authority on OpenAI, Keach Hagey notes, perhaps the most hallucinatory feature the $300 billion start-up co-founded by the deadly duo of Sam Altman and Elon Musk is its attempt to be simultaneously a for-profit and non-profit company. As Hagey notes, the double life of this double company reached a surreal climax this week when Altman announced that OpenAI was abandoning its promised for-profit conversion. So what, I asked Hagey, are the implications of this corporate volte-face for investors who have poured billions of real dollars into the non-profit in order to make a profit? Will they be Waiting For Godot to get their returns?

As Hagey - whose excellent biography of Altman, The Optimist, is out in a couple of weeks - explains, this might be the story of the hubristic 2020’s. She speaks of Altman’s astonishingly (even for Silicon Valley) hubris in believing that he can get away with the alchemic conceit of inventing a multi trillion dollar for-profit non-profit company. Yes, you can be half-pregnant, Sam is promising us. But, as she warns, at some point this will be exposed as fantasy. The consequences might not exactly be another Enron or FTX, but it will have ramifications way beyond beyond Silicon Valley. What will happen, for example, if future investors aren’t convinced by Altman’s fantasy and OpenAI runs out of cash? Hagey suggests that the OpenAI story may ultimately become a political drama in which a MAGA President will be forced to bail out America’s leading AI company. It’s TikTok in reverse (imagine if Chinese investors try to acquire OpenAI).

Startup of the Week

Lightricks just made AI video generation 30x faster — and you won’t need a $10,000 GPU

Venturebeat • May 6, 2025

Technology•AI•MachineLearning•VideoGeneration•OpenSource•Innovation•Startup of the Week


Lightricks has unveiled its latest AI video generation model, the LTXV-13B, which represents a significant leap forward in both quality and efficiency for AI-driven content creation. The new model, boasting 13 billion parameters, is designed to deliver high-fidelity video output while maintaining the speed that made its predecessor, the original LTXV, a standout in the field. Notably, LTXV-13B can generate videos with "stunning detail, coherence, and control," even when running on consumer-grade hardware, making advanced AI video accessible without the need for expensive, high-end GPUs.

At the heart of this breakthrough is Lightricks' innovative "multiscale rendering" technology. This approach allows the model to draft video content in lower detail first, capturing coarse motion and structure with minimal computational resources. The initial draft then guides subsequent stages, where the model progressively refines the video by adding structure, lighting, and micro-motion. By focusing computational effort where it matters most, LTXV-13B achieves render times that are more than 30 times faster than comparable models, all without compromising visual realism.

The LTXV-13B model is available within Lightricks' flagship storytelling platform, LTX Studio, and is being integrated across the company's product portfolio. Its open-source nature sets it apart from many proprietary models, offering researchers and enthusiasts the opportunity to understand, modify, and improve the technology. This accessibility is a key differentiator in a landscape where advanced AI models are often locked behind pay-to-play APIs or require specialized, costly hardware.

Lightricks' commitment to leveraging the latest advancements from academia and the open-source community is evident in LTXV-13B's architecture. The model incorporates unsampling controls, spatiotemporal guidance for video editing, and kernel optimization for running speeds, further enhancing its performance and usability.

With this release, Lightricks is positioning itself as a major player in the AI video generation space, challenging established tech giants and democratizing access to cutting-edge creative tools. The LTXV-13B model promises to empower creators with unprecedented speed, quality, and control, all while running efficiently on everyday hardware.

Read More

Post of the Week

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.

Discussion about this video