We don’t really know what AI could become…

Good Morning,

The two biggest events this week are likely the reaction to Fable and the SpaceX IPO.

I’ve been struggling a little trying to think about what’s really new that was announced at Google I/O, Microsoft Build Conference and Apple’s own more recent developer conference of WWDC 2026. There was a tangible sense of how empty many of their tech announcements were, especially with regards to AI.

I can’t even put my finger on why this is, maybe it’s the AI overload of the news cycle, the exaggerated hype or how lacking in meaningful products and application layer innovations we’ve even seen after three years of LLMs (so many models) and more datacenters. I can’t help put have the feeling that I was expecting more in the 2020s that we are likely to see from Generative AI.

Hopefully next year in 2027 with a deluge of AI devices this will feel different. Will a Codex version of ChatGPT feel different? Apparently, a senior OpenAI executive went so far as to say that “chat is dead,” shifting its focus toward a platform centered on agents and coding tools. It just goes to show how ChatGPT’s success story is fading so badly. But how would OpenAI ever compete with Cursor or Anthropic at this point for Enterprises and companies where trust is a major factor. OpenAI might have to try to lower prices, but they are already burning so much capital. Cursor who now have $4 Bn. in ARR and are accelerating even faster.

There certainly will be winners and losers to paraphrase Sam Altman. His eye-scanning startup isn’t fairing any better.

Will BigAI Disrupt Incumbents by 2030?

Will the hyperscaler names be re-written by AI too? 🤔 I had written about this back in June, 2025:

The verdict on how BigTech (incumbents) react to BigAI is still unclear? Outside of Google, it doesn’t look so good for the others. Their capex to AI execution is looking rather poor in more than one case.

Let’s first see how the 2026 IPOs pan out but Apple, Microsoft, Amazon and other incumbents don’t feel like they are active participants at the frontier of Generative AI. Apple’s lineup of AI devices looks fairly impressive however for 2027 and 2028.

China’s Massive AI Compute Announcement

With news that DeepSeek is getting $7 Billion in funding a week ago along with massive Capex increases from ByteDance and Alibaba for compute - China is going all in. China is preparing to spend around 2 trillion yuan ($295.43 billion) over the next five years on building data centers ​across the country, Bloomberg News reported on Tuesday, as ‌Beijing looks to challenge the U.S. in the intensifying AI race.

So instead of relying purely on private tech giants like Alibaba, ByteDance or Tencent to build isolated clouds, the National Development and Reform Commission (NDRC) is spearheading a blueprint to knit the country's scattered computing facilities into a unified national network by 2028. If China feels they have to do this too, it’s going to be a brave new world. Curiously how Beijing is boosting Huawei isn’t so different how the Trump Administration has been supporting Nvidia with circular financing of the entire ecosystem in a spider-web of deals. But what exactly are we building for civilization in the big picture.

We Don’t really know what we are getting ourselves into…

Claude Skills Guides

• KDNuggets (not an affiliate) has an interesting Anthropic’s Complete Guide to Claude Skills Building.

• The Claude (Content) skills I use to run Growth Unhinged by

• How to use Claude AI for Investing (the first post in our series on AI in Investing got a lot of bookmark saves on X).

Reactions to the Frontier Model of Anthropic

Reactions on X to the Fable 5 release were conflicted and fairly heated: 🔥 Also baffling to me considering the pressure Anthropic is under both from a legal, compute constraint and pre-IPO perspective.

Fable 5 is Controversial

I wanted to gather some quotes to give you a live-sentiment sort of reaction to the announcement and nature of Claude Fable 5.

Quotes:

“There is a lot of justified anger at Anthropic for sandbagging Fable 5 for AI development tasks. But an unanticipated side effect is that third-party evaluators can no longer credibly use the model for evaluations.” - Sayash Kapoor

“In experiment after experiment I conducted, it outperformed basically every other public model I have used by a considerable margin. It was capable across many problems and produced some startling results — it would work up to a dozen hours executing on multi-page specifications.” - Ethan Mollick

“Anthropic designing Mythos/Fable to prevent frontier AI research is the most disappointing part of their release. Hope it doesn't set a precedent (I don't think it will).” - Ben Dickson

“Claude Fable 5 is by far the most ridiculous model that makes me genuinely afraid for the future of software engineering.” - Deedy Das (Menlo Ventures who are investors in Anthropic).

“Claude Fable 5 just changed the AI game. People are one-shotting games, 3D worlds, app builders and insane code optimizations. There's a major shift.” - Min Choi

“When Fable 5 is used for frontier LLM development, it does not notify the user and instead limits the model’s capabilities through methods such as prompt modification, steering vectors, and PEFT.” - User @Hangsiin - this is thought to be the account of Seungwoo Jeong.

“Mythos / Fable is unbelievable. Was on a customer call today and had Claude transcribing in the background. As they were telling me about the features they wish their current software had, Claude was building the features in real time.” - Todd Saunders

“Claude Fable 5 is likely very capable inherently on healthcare. That's great! Too bad it's near impossible to tap into those capabilities due to their extremely sensitive safety filters. I hope this is adjusted going forward.” - Tanishq Mathew Abraham

Things I really dislike about Fable:

1. Anthropic collects my prompt history, stores it, and does whatever they want with it for 30 days. No opt-out

2. They can nerf their most expensive model without telling me, billing me the same amount, wasting my time. Whenever they want. - Gergely Orosz

“The fury is real and what all of us in the open community have been saying for years and yet regular folks don't get it yet because nothing they care about is restricted or taken away for "safety."

They will care a LOT in the future when AI is integrated into every aspect of our lives. When it's our interface to the world and knows every intimate detail about us.” - Daniel Jeffries

“Labs starting to pull up the ladders on the ability to diffuse AI was inevitable. Doing it without telling the user is misaligned.” - Nathan Lambert

“Degrading performance on ML research *without telling the user* is shockingly hostile and a terrible look. That could silently damage all sorts of work, including some of my own. Also the type of thing that could raise the eyebrows of antitrust enforcers worldwide.” - Dean W. Ball

“Fable is the best model I have used for coding, by a wide margin. It is a big step up, enabling less prompts and steers, more efficient token use, better code quality, better tool use, more intelligent self-verification, longer running sessions, and higher trust & autonomy.” - Boris Cherny

“Anthropic wants to control who gets access to their models and what they're allowed to do with them, but also wants the US government to block Chinese labs from developing open weight models. Sorry, but ***[no] (edited for profanity). - Zach Korman

“The scary part about Anthropic's Fable nerf is not that it refuses to answer biology or cryptography. It's that it foreshadows what's coming. A world where a couple companies decide what you can and cannot do. They're building a new ruling class and you're not in it...” - Péter Szilágyi

My SpaceX IPO analysis at the end. The intersection of leadership, AI policy, Government and risk-management is not looking good in 2026. America’s commitment to closed-model AI Supremacy is not looking benevolent either.

Watered Down Mythos 🌊

Anthropic remains compute constrained and in legal disputes with the Pentagon, and their latest model clearly suffers from both. Many AI researchers are not happy with the Fable 5 release. They are five months from an IPO and are the global leader in Coding AI models and likely Agentic AI execution. That earns them the right to do what they think is right. But preaching about slowing down or more regulatory oversight required? It’s way too late for that.

Anthropic’s simultaneous release of Claude Fable 5 and Claude Mythos 5 marks the public debut of their next-generation, high-autonomy "Mythos-class" architecture, but it’s also not very popular and misunderstood.

• Fable 5 and Mythos 5 are being offered at $10 per million input tokens and $50 per million output tokens—less than half the price of Claude Mythos Preview.

• Anthropic says that they intend to expand access to Mythos 5 through a broader trusted access program, both for defensive cybersecurity work and biomedical research.

• The safeguards, which mean some queries on certain topics will instead be ported to Anthropic's Opus 4.8 model, are initially tuned to be more conservative. The safeguards are triggered when Fable 5 detects requests concerning cybersecurity, biology, chemistry and (model) distillation. Anthropic seems to be trying to build a moat via rather intense safeguards.

• Anthropic’s CEO recently published yet another essay:

Read the Essay

He essentially is calling for more Government regulation which is a bit hypocritical since his AI company is the one that’s currently ahead. He argues AI capabilities are accelerating at an exponential pace, creating an urgent mismatch with the traditionally slow, deliberative nature of government policy and legislation. However AI researchers point out that restrictions on Fable 5 are also too restrictive and non-consensual. Many after the Fable 5 release expressed renewed skepticism in Anthropic’s image as being a trust & saftey lab that’s simultaneously the leader pushing the frontier of AI at the B2B and Enterprise product level. Its incentives are no longer aligned.

“Any reasonable analysis of the past five years makes it abundantly clear that major AI companies are making the world a worse place and open source AI isn’t. So they distract you from the actual harm they’re doing in the world with hypothetical fearmongering about the future.” - Stella Biderman

By projecting itself in the future and being fairly ideological, political and with a CEO who makes predictions/warnings on multiple occasions making statements/claims that sound like fear-mongering, Anthropic has lost considerable support in the community. This was one of the biggest takeaways this week in AI for me. Unfortunately this lack of leadership starts at the top. AI is bound to get more political in such a climate while being inflationary in the midst of an affordability crisis. This is also lowering the sentiment of the public towards AI.

We’re Missing the Boat on AI with a lack of Leadership ⛵

0:33 seconds. I really like this quote: Andrew Yang said that an absence in leadership has made it so the people are figuring out how to handle AI-related issues.

Anthropic’s Mythos Models are a Noticeable Step-up

Anthropic insists Fable/Mythos models are frontier especially in the following:

1. Software engineering

2. Knowledge work

3. Memory and long-context

4. Vision

SpaceX IPO Goes live on Friday

Tomorrow is the big day. But will the people figure out SpaceX and its issues? Heading into its abbreviated roadshow last week, the company said $135 is the share price and $1.77 trillion is the anticipated market cap.

I have been reading a lot of SpaceX IPO articles and some of the infographics really caught my attention. Let me share some I’ve found along the way:

The Price to Sales Ratio of SpaceX

IPOs will dump roughly $300 billion of new stock in the market

BigAI is going to raise nearly $300 Bn.

Space, Tesla and xAI Operate practically as one Entity

In an era of rampant circular financing, Elon Musk’s companies already operate in many self-reinforcing loops.

Orbital datacenters, ramping more Starship launches, TeraFab and scaling xAI will cost way more capital and Capex than SpaceX & Tesla possess. We will approach a time in the 2030s when capital & debt limits are reached. SpaceX doesn’t have much of a window.

Buying Most IPOs when they go public is a bad deal

Outside of Palantir, Arm Holdings, Snowflake and Airbnb, most IPOs are over-priced. Down on average 53% one year later. The SpaceX valuation or the TAM it claims in its S-1 is beyond possible. An older Elon Musk is not better than what he was in his prime, paper Trillionaire or not.

The Valuation is most like a Gigantic Ponzi-Scheme

SpaceX took 2 decades to cross a $200 billion valuation. It added the next $1.5 trillion in 18 months. Does that sound rational to you? The only meaningful revenue it’s now generating is leasing out compute to the likes of Anthropic and Google. Alphabet that owns a sizeable stake in SpaceX and also a sizeable stake in Anthropic.

SpaceX is valued as Big as all of Aerospace Stocks

It better deliver some National defense relevance with a valuation like that. 🤔

Will SpaceX even Acquire Anyspherer (Cursor)

xAI could scale, but most likely only if it was able to acquire Cursor, that has now reached $4 Bn, in ARR. SpaceX holds the exclusive right to fully acquire Cursor for $60 billion. The plan is to finalize this roughly 30 days after SpaceX wraps up its blockbuster public listing. That doesn’t mean it’s likely to happen. xAI has failed to achieve either B2B and B2B scale in terms of revenue, inspite of a lot of debt and burning huge amounts of capital. SpaceX isn’t really an AI company so far as meaningful revenue is concerned at least in 2026. The dangers to the public are also real.

SpaceX Hasn’t Scaled Revenue well Historically

It’s only renting out their compute to Anthropic and Google that saves their ARR projected in 2026 to be considerably higher than 2025. SpaceX is thus more of a Neo Cloud today, than a rocket company or a connectivity company.

Prior to these infrastructure deals, SpaceX’s baseline 2026 core revenue (Launch operations + Starlink subscriptions) was projected to hit just $22.0 billion. But the compute deals lifts its 2026 expected revenue to about $34 Bn. That still doesn’t explain a valuation of nearly $2 Tr. Blue Origin’s rocket blowing up was maybe the best news for the IPO. Some Bezos didn't have a contingency plan. Their Launch Complex 36 (LC-36) in Florida was totalled setting back their Starlink competitor significantly.

Will SpaceX just be a Neo Cloud?

The only area where you could legitimately think of xAI or SpaceX at the frontier is their investment in compute (see image below). So to make ends meet they might just have to become a Neo Cloud. If they merged with Tesla at least they would have the Energy and battery storage business. It takes revenue and capital for SpaceX to scale their moonshots, money they don’t have.

As disclosed in its recent S-1 registration filing ahead of its public debut, SpaceX holds approximately $29.1 billion to $30 billion in total debt as of the end of the first quarter of 2026. Crucially, $20 billion of SpaceX's debt is structured as a short-term bridge loan. A lot of the proceeds from the IPO just goes to paying off debt. Their $75 Bn. from IPO proceeds doesn’t go very far.

SpaceX spent a staggering $7.7 billion on AI-related CapEx in Q1 2026 alone. xAI is burning cash as if it was making ARR like Anthropic or OpenAI, but it’s not. That’s a problem. Other rocket companies like Rocket Lab are quickly catching up to its in the ability to do many rocket launches. More rocket companies are coming. That’s not counting China capabilities in rocket launches and space-infrastructure.

Time will tell if SpaceX will be propped up with the (fast-tracked) early entry into the NASDAQ. It has its work cut out for it as Elon Musk’s companies have to pivot in a world where they are or won’t soon be the first mover winners in much of anything.

Thanks for reading!

As SpaceX, OpenAI and Anthropic IPOs approach in the days and months ahead, how we define the narrative around the future of AI is also taking shape. With Anthropic expanding access to its new class of LLM called Mythos to more organizations, the U.S. Department of War appears to be using Mythos for Cybersecurity attacks, reports the FT.

In recent weeks several significant seed rounds have been given to AI labs in the U.S. working on various aspects of recursive self-improving AI capabilities, or RSI. We have to conclude that Mythos as an LLM step-up might represent the first RSI-native Model. If Anthropic’s framing is correct, we are heading into unchartered territory.

Claude 4.8 already seemed like solid advancements but the Mythos class appears relatively distinguished compared to its competitors at Google, OpenAI or China. This is giving Anthropic a feeling of grandeur heading to its landmark IPO that might end up being the highlight of these Mega IPOs of 2026. Now in a blog post on its website on Thursday, Anthropic wrote about the risks of AI progressing to a point where it could improve itself autonomously.

Anthropic reported a landmark blog titled When AI Builds Itself, while the U.S. Government is discussion about taking equity stakes in the biggest AI companies. That is for real, Senior U.S. officials held preliminary discussions with major AI companies about the potential for the government to buy some shares in their firms. These are unprecedented and gravely risky endeavors as Nation States seek to mould AI in their image.

OpenAI CEO Sam Altman has discussed the idea with government officials since President Donald Trump began his ​second term, the report said. Sam Altman and Bernie Sanders have different ideas on how that could look. This is in sync with Trump and his financial elite friends to create a Sovereign Wealth Fund for the United States. Anthropic is not having conversations with the administration about providing equity to ​the government and it’s not clear what Sam Altman or OpenAI’s motives are to enter into such a deal.

A White House meeting is reportedly being set up with leading AI firms to explore the concept

Anthropic’s Mythos has raised concerns that new AI models might be capable of, including and especially in cybersecurity. Anthropic confidentially filed its IPO prospectus with the SEC and said it will offer Mythos to the European Union along with an additional 150 partners in 15 countries. Mythos is becoming more than just a frontier model however, it’s becoming part of a larger debate around what recursive self-improving AI might do if it’s real across industries.

Anthropic’s cryptic Tweets are also unsettling at a time when Anthropic is clearly working more closely with the Department of War while being held semi-hostage with a rather nefarious designation as a supply-chain risk.

Defense Secretary Pete Hegseth has denied Anthropic’s request to reconsider the AI startup’s designation as a national security risk, the Pentagon told the D.C. Circuit Court of Appeals on Thursday.

It appears like the Trump Administration has a special relationship with Sam Altman and OpenAI and is trying to supress whatever Anthropic represents. If Mythos is being used for cybersecurity offense (supposedly against China) then this is not a good sign of AI becoming a geopolitical risk in an era of more automated systems in National defense.

Anthropic, is Everything Alright?

Read more

Good Morning, everyone.

It’s a very hot and humid June here near Taichung, Taiwan.

Welcome back to AI Supremacy where we cover AI at the intersection of business, finance, tech, society, culture and civilization. We also try to deliver to you a few AI guides each month.

I’m not sure if Anthropic is 'midwifing a deity' according to VC Bill Gurley, but they have seemingly beaten OpenAI to a confidential IPO filing that enables them to fast-track what will likely be the biggest AI IPO in history in September, or October later this year. It’s a frantic situation for some of the fastest AI startups in history.

“The AI supercycle has 4 bottlenecks: compute, memory, power and now photonics.” - Crack the Market Newsletter ()

It’s an exceptional time, Google wants to raise another $80 billion in stock, including through a $10 billion investment by Berkshire Hathaway for the unique challenges of more AI Infrastructure and power solutions.

Claude for Equity Research

Meanwhile our guide on using Claude for investing by got a lot unexpected organic traffic and bookmarks on X, where people are using AI for equity research, and beginning to use agents for investing. What struck me about this piece is the number of bookmarks, free trials and buzz it was getting across social media.

Read our Guide

The Datacenter Acceleration & Backlash 🚨

There are a lot of Datacenter stories that I want to catch up to drive and give you a holistic picture from many angles about the space and the reality not just from the VC and corporate side but from the more human, energy, local community and environmental perspectives as well.

In April, Alphabet updated its full-year capital expenditure range to as much as $190 billion and others like Meta and Amazon may go even further so there’s a steep increase in corporate foreign bonds (debt) and various projects around Energy and Datacenters. With SpaceX having to sell compute to Anthropic just to make ends meet pre-IPO, Meta might become a glorified Neo Cloud as a side gig. These my friends are weird times. Google had already last month announced a TPU based Neo Cloud with Blackstone.

In this article I’ll try to cover some trends I’m noticing and some Datacenter stories that caught my eye over the past few weeks. Scroll down for more on this.

AI Startups are Eating Pre Generative AI Startups: Fallen Unicorns 🦄

In Venture Capital’s obsession with AI, they are funding newer startups that are AI-native post 2023. For instance, AI is “disrupting” startups pre ChatGPT. Nearly half of America’s 857 unicorn startups haven’t raised fresh funding in three years, PitchBook data shows.

While business creation appears to be higher, there’s carnage in the startup world. Nobody told you more than 220 companies that once hit billion-dollar valuations are now considered “fallen unicorns”.

The concentration of capital is also damaging - in that, the AI boom that has funneled more than $250 billion into OpenAI and Anthropic and reset valuations on entire classes of startups. There’s a kind of price for this kind of American innovation. It’s pure carnage in the land of startups and Silicon Valley doesn’t like to admit the obvious:

“Startups that last raised in 2021 are now worth 68% less on average, while those that last raised in 2022 saw a 52% decline, according to Pitchbook’s own valuation estimates.” - CNBC

Startups that raised just 4 years ago, in 2022, are down 52%. Just like I predict now, startups born with agentic AI might destroy AI startups pre Agentic AI, for example pre 2026. AI in many ways is a great destroyer of capital not just in Capex but in the cost of Venture Capital FOMO on all the other industries that would ordinarily be getting funded.

It’s not just the demand for compute that’s accelerating, it’s the speed of adaptation.

Time to Revenue has Been Sped Up

Read more

Read more

The antics of AGI of 2023 is going to turn into the race to recursive AI (RSI) of 2027. Recursive self-improving AI could fundamentally reshape how we experience the world and fundamentally accelerate digital transformation in the years to come.

While Anthropic has been claiming they use Claude for the majority of their coding since quite a while, there are no signs that we are on the cup of RSI just yet in mid 2026.

However there are a surprising number of new AI startups that claim they are building recursive capabilities of AI into super valuable industries, like chip design. Many of them claim this is their academic focus now and the startup called Recursive Superintelligence really comes to mind most recently. You can imagine how AI engineering their own improvements can scale fairly quickly once and if it gets good at this.

But will this propel a new era of science, R&D and innovation broadly speaking? And what are the chances of these buzzy startups at being commercially successful and actually succeeding in a commercially viable way at what they propose will soon be possible? Let alone with just Generative AI technology.

“At its core, recursive self-improvement (RSI) is a process where an artificial intelligence system reviews, rewrites, and optimizes its own source code or architecture to become smarter.”

• I believe 2027 will indeed feel like the entry point into a fundamentally different world post the three major AI IPOs of 2026. The vibe-shift in the AI ecosystem is going to be tangible.

• DeepSeek’s senior management has told potential investors in its ongoing 70 billion yuan ($10 billion) funding round that the startup will prioritize groundbreaking AI research over short-term commercialization.

• With AGI not a commercial benchmark worth talking about seriously, RSI becomes a much more concrete proposition. The idea of AI improving itself exponentially is a fundamental inflection point in the theory around the Technological singularity. Ray Kurzweil, in his book The Singularity is Near, (that I used to obsessed over many years ago) famously predicted that the singularity will occur by 2045. Now some believe that it could arrive anywhere between the years 2030 and 2045. Some see the rapid improvements of Generative AI’s capabilities as evidence it is indeed on the way to happening.

From AGI to RSI Goals for the Future of AI

"Today's systems are nowhere near [AGI]. Doesn't matter how many Erdős problems you solve… I think it's far, far from what a true invention, or someone like Ramanujan, would have been able to do." - Demis Hassabis, CEO of Google DeepMind and Nobel laureate for AlphaFold

Context: Demis Hassabis isn’t just Google DeepMind’s Czar, it turns out he’s an angel investor in Anthropic. It’s a tightly woven interconnected web of influence - where the network of startups spun out by DeepMind alumni have collectively raised a collective $14 billion plus since 2021.

Now the same is occurring with OpenAI alum.

Hassabis invested personally in the early/angel stage, before Anthropic became a major player valued in the tens or hundreds of billions, and I’d estimate could have up to a $2 Billion personal stake. These financial conflicts of interest are becoming quite common in the AI ecosystem especially clustered around OpenAI, Google DeepMind and Anthropic and their Alum. AI executives increasingly are angel investors on the side.

“People familiar with the matter said Hassabis was an angel investor in Anthropic, which has since become a key AI rival and among the world’s fastest-growing start-ups. Google has separately invested billions of dollars in Anthropic as part of its cloud and AI partnership with the company.” - The FT

How much of an impact was Hassabis on Alphabet itself investing huge sums into Anthropic and using their TPUs do you suppose? Or decisions to partner with Blackstone on new Neo Cloud do you suppose? These are newly available factors in the circular financing argument.

Manufactured Geniuses hidden in Rising demand for Compute

Read more

Good Morning,

I’m still digesting the Google I/O conference news and some of the key AI trends I’m noticing and while I do so, I wanted to share something a bit different from the usual. It’s a potentially really useful guide since we are in an AI boom of unprecedented opportunity and stock market euphoria that is perhaps a unique window.

Michael Fritzell is an independent financial analyst, investor, and writer based in Singapore. He is best known as the founder and author of Asian Century Stocks, a popular investment newsletter. Michael is a friend of mine and one of the most successful financial writers I know who grew his Newsletter on Substack pretty early on.

This is going to be a guest post by Michael Fritzell who runs one of the most useful investing Newsletters I know of especially for Asian equities. I wanted to share this practical guide since how investors and analysts are using AI in Finance is changing very fast. This is a hands-on guide that should give you a lot of ideas on where to start to use Claude for your investing decision making.

Sign up to Asia Century Stocks

Michael’s Newsletter is no longer on Substack but moved to the Ghost platform. A lot of investors are of course using Claude for their investing activities.

The author briefly describes his Newsletter publication: 1 min 28 seconds. 🎬

• You can find more of his recent interviews with Substack investing, FinTwit and equity folk here.

Many of my readers are also investors and think about technology and businesses at a very deep level.

• Read his recent Deep Dives.

• Follow Michael Fritzell’s Portfolio.

• View his Thematic Reports

If you want to dig deeper to get full access to the work of someone that’s been studying Asian equities for nearly two decades, consider a full sub to Asian Century Stocks.

Get full Access to ACS

One way to think of using Claude for Equity Research is the following:

1. Research

2. Projects

3. Skills

4. Cowork

Claude has evolved a lot when it comes to Agents for Financial Services. Download Claude Cowork.

Some Starting Ideas

• Investors instruct Claude Code to scan the web, pull down the latest 10-K forms, SEC filings, and earnings call transcripts directly into local directories.

• Live portfolio analysis: with Anthropic's Model Context Protocol (MCP), users are plugging Claude directly into live financial databases and brokerages.

Claude for Equity Research

By

Biographical Sketch 🏛️💰💸💳💶

• Originally from Sweden, Fritzell started his career in investment banking in the United Kingdom after completing his university degree.

• After studying Chinese, he moved to Shanghai to work as a buy-side analyst for an emerging markets fund, traveling extensively across China to evaluate investment opportunities and research local companies.

• In 2014, he relocated to Singapore to manage an Asian equity portfolio for a European family office, expanding his coverage to markets like Indonesia, Thailand, Vietnam, South Korea, and the Philippines. He later worked for a hedge fund and a mutual fund in Singapore.

• In April 2021, Fritzell left the traditional fund management industry to write Asian Century Stocks full-time. He’s now based in Singapore.

Michael is a professional analyst, he always includes a disclaimer so the rest of the article is by him:

Disclaimer: This article constitutes the author’s personal views and is for entertainment and educational purposes only. It is not to be construed as financial advice in any shape or form. Please do your own research and seek your own advice from a qualified financial advisor. From time to time, the author might hold positions in the below-mentioned stocks consistent with the views and opinions expressed in this article. This is a disclosure, not a recommendation to buy or sell stocks.

2026 has been the year of Anthropic.

They are the creator of the large language model Claude, which has taken the world by storm:

The company was founded by a group of ex-OpenAI executives, who purportedly set out to create a more ethical generative AI tool. The result was Claude.ai – a chatbot that functions much the same way as ChatGPT and Google Gemini.

However, the popularity of Claude did not take off until late January 2026. At the time, Anthropic released Claude Cowork — an AI colleague that can work with files on your computer using natural language prompts.

And since then, Anthropic has been on an epic run of product releases, including Claude Design, Claude Marketplace, Claude for Excel, Claude for PowerPoint, Claude for Word, Claude for Chrome, and Claude Mythos.

In this post, I’ll explain exactly how I use Claude to analyze stocks. And hopefully, you’ll learn something in the process.

Personalized instructions

First, go to claude.ai and create an account.

There’s a free plan that gives you access to the basic model (Claude Sonnet) with up to 100 messages per day.

There’s also a Pro plan for US$20/month that gives you access to the reasoning model Opus 4.7, with 5x more usage than the free tier, access to Claude for Microsoft Office, and Claude Cowork. I suggest buying the Pro plan.

When you first go to the website, Claude won’t know much about you. So I suggest going to your profile (in my case, “MF”) and then “Settings”. You’ll now be able to provide context on what you’re looking for from Claude.

In my case, I tell Claude that:

“I am Michael Fritzell, an equity research analyst. Whenever I ask about a particular company, I want to understand the fundamentals of the business model – in other words, how it makes money. I also want to understand what makes the company unique in comparison to its competitors. Be brief and to the point.”

If you’ve used Claude for a while, it will have saved a memory of past discussions. You can access that memory by clicking “Memory from your chats”:

If you want to add anything to the memory, click the pen button at the lower-left corner of the screen and add things you’ll want Claude to keep in mind.

You can also import the memory from other generative AI tools. Just click “Settings” and then “Capabilities”, and follow the instructions after clicking “Start Import”.

Claude will now have much greater context on who you are and what you’re looking for.

Managing the main prompt window

While visiting the website, you can reach Claude’s main prompt window by clicking “Ctrl/⌘ + Shift + O”. You’ll then be presented with the following window:

To add files, click the “+” sign, type “/” or drag and drop a file from your computer. If it’s a picture, you can also just paste it by typing “Ctrl/⌘ + V”.

Use Sonnet 4.6 for routine questions and tasks and Opus 4.7 for complex, multi-step analysis (if you’re on the Pro plan).

If you click the sound button in the lower-right corner of the screen, you’ll get to Claude’s voice mode, where you can talk to the AI model using your computer’s microphone. I’ve found Claude’s live mode better than Gemini’s but worse than ChatGPT’s. See what you think.

Save AI prompts in Claude Projects

Read more

Good Morning,

AI Supremacy is a Newsletter that explores the impact of AI on the future of business, technology, culture and society.

This is starting ot mean a lot for the lives of different kinds of people, consumers, investors, graduates, entrepreneurs, citizens and residents. The pressure of AI on real people and their communities is becoming more tangible this year.

A Pivotal AI Week Ahead

With Google’s I/O Conference starting tomorrow, Nvidia Earnings and the SpaceX IPO filing, this might be one of the most pivotal weeks in AI of 2026 (week of May 18th). Elon Musk is expected to start meeting with investors very soon ahead of an IPO for SpaceX, which was valued at $1.25 trillion after merging with xAI in February. I’m noticing a lot of subplots in the AI ecosystem in a rather bizarre geopolitical, inflation, bond story and Silicon Valley backdrop. Let’s dive into some of the most important among these.

Earnings Calendar this Week

Last week Anthropic announced Claude for Small Business, less than a year now from its IPO likely in December, 2026. You hear a lot of stories of SMBs struggling to implement Generative AI technology. This is especially true in an era of rising backlash against AI slop, datacenters and Billionaire incentives to praise the AI era as if it’s going to save us all.

At the time of writing today, Cursor released Composer 2.5, and I believe after Anthropic, Anysphere is one of the top B2B and Enterprise AI startups. From a product perspective I think this improves or at least enhances Cursor’s in-house model intelligence, long-running agentic behaviors, and the overall team development ecosystem within its IDE. Cursor is becoming known for its ability to multitask its strong parallel processing with async subagents.

The midpoint of 2026 marks an extremely cringe time in the history of AI, where entry level jobs are being impacted, and people are already extremely fed up. According to Axios, the “disdain spans generations” and political parties. An Economist/YouGov poll released this week showed over 70% of Americans think AI is advancing too quickly, with 68% of Republicans and 77% of Democrats saying it’s moving too fast.

The American Billionaire AI Rush 2026-

Capital, compute and the AI boom of elitism. What’s taking shape is not necessarily making many of us more hopeful about the future.

With three huge IPOs, SpaceX, Anthropic and OpenAI supposedly all coming in 2026 or early 2027, there’s a lot of capital at stake for a lot of rich people and Venture Capital funds, BigTech and the perception that America is ahead in AI.

A jury in Oakland, California, ruled against Elon Musk in his dramatic court battle with Sam Altman and OpenAI. Sam Altman is among a new breed of AI Venture Capitalists that are powerful angel investors while assuming CEO duties of one of the most powerful AI startups ever in OpenAI. I’ve been arguing that it is a company in decline. Whatever it is, it’s a child of a new era in Venture Capital history dominated by a few big funds.

Venture Capital Power is becoming more Concentrated

The Q2 2026 ranking of US unicorns shows a rather high concentration of VCs involved. They are roughly as follows:

A small handful of VCs are seeing the majority of the returns in the AI boom to the detriment of mid-sized VCs. Generative AI is narrowing the power and concentration base of Silicon Valley that was already previously highly concentrated.

AI is creating fewer winners in VC, BigTech and the Future

GenZ is waking up to the danger of BigTech in 2026

Read more

Good Morning,

Innovation in AI is occuring at many of the latest AI Research labs. This has huge ramifications for the future of AI.

What if AI models could deliver experiences that were more human like? As an avid user of Voice AI, I’ve often wondered about this. This week we finally got to see what Thinky is up to.

Thinking Machines Lab, the AI startup founded last year by former OpenAI CTO Mira Murati, that has weirdly enough been mercilessly poached by OpenAI, and yet they have been able to attract some of Silicon Valley’s top AI talent. Thinky wants to move beyond the era of “turn-based” AI interactions and, I think they are on to something.

If they can pull it off, this will essentially deliver way more real and human-like AI interactions in real-time. Think about it, traditional AI models function sequentially—listening and then responding—whereas its new model aims for simultaneous input processing and response generation, akin to a phone call.

What are Interaction Models?

While Lilian is telling a story, the interaction model can track when she is thinking, yielding, self-correcting, or inviting a response; there is no specific built dialogue management system.

Lilian Weng is a Co-founder at Thinky.

Interaction Models are a new category of AI systems with micro-turn architecture, a system that processes data in 200ms chunks. This could literally make AI interactions a lot more useful in the real physical world and social world we actually live in. That’s more human-like for sure Thinky! Murati is no deer in the headlights, she’s just Albanian and the technical term for this is “full duplex,” and the company claims its model, TML-Interaction-Small, responds in 0.40 seconds, which is roughly the speed of natural human conversation and significantly faster than comparable models from OpenAI and Google.

I really dialed into this story since this will clearly lead to a lot of other interesting AI products. If you are a builder, developer, investor or power-user of AI - this is incredibly exciting.

A More Natural AI Interface is Emerging

Essentially AI should serve us but since 2023 we’ve been forced into bad interactions. There’s still a long ways to go. Thinking Machines argues that the back-and-forth interactions with current models forces human users to “contort themselves” to the interface. If you are a regular Voice AI user you know exactly what I’m talking about.

Could we have Benevolent AI that actually looks out for us?

Instead of AI trying to scam us with pleasing words?

Tessa's quality of life has improved a lot with some nagging.

I actually found this pretty funny and would be fine with an AI that nags me in a good way. Can we envision a time when AI is a benevolent advisor and implementer across human institutions and helping to guide humanity across civilization scale challenges? Perhaps like me you are familiar with how the concept of an artificial intelligence "Oversoul" is famously explored by author Orson Scott Card in his 5-book science fiction series, the Homecoming Saga.

For me, Interaction Models point to that world perhaps decades away.

So what would it be like to interact with an AI that isn’t so fake and contrived? The Interaction models have real potential to humanize AI in a way that I think is fairly positive. The interaction models is essentially a new AI system built to enable real-time, native collaboration across audio, video, and text without relying on external scaffolding. You an almost feel the feminine touch in Thinky’s product focus. AI doesn’t have to be awful and feel flawed and sycophantic. Is there a better way to represent and build products with this technology?

‘Full duplex’ simultaneous input/output processing

Thinking Machines is essentially here helping to solve the "collaboration bottleneck” in how we relate with AI in a more realistic manner.

Read the Blog

“We think interactivity should scale alongside intelligence; the way we work with AI should not be treated as an afterthought. Interaction models let people collaborate with AI the way we naturally collaborate with each other—they continuously take in audio, video, and text, and think, respond, and act in real time.” - Thinking Machines Blog.

Capabilities of Interaction Models

Read more

Good Morning,

I’ve been following the Elon Musk trial vs. OpenAI and Sam Altman and there’s reason to believe OpenAI is on a downward spiral. To put it bluntly, OpenAI’s momentum has now been overtaken by that of Anthropic, Google and other AI labs. I’m not sure how it rebounds from this. In terms of LLMs and AI products this is how I see the ecosystem and companies with positive momentum and who are falling behind.

It took me many hours and many days to compile this, so I hope you enjoy it. Pre IPO for the biggest three AI IPOs in history in Anthropic, SpaceX and OpenA, it’s sort of a big deal.

Who’s Hot 🔥

• Anthropic

• Anysphere (Cursor)

• Google

• DeepSeek: aiming to raise up to $7.35 billion (50 billion yuan) at a valuation potentially reaching $50 billion.

• Moonshot AI: $2 billion in a new funding round.

Who is Not 🤯

• OpenAI: credibility loss, investors doubts and scaling back compute

• Microsoft: lagging in internal models and reputation

• Meta: poor execution and exodus of talent

• Amazon

• Apple

OpenAI’s has Executed Poorly and No longer has a material Compute Advantage

OpenAI’s leadership credibility and reputation is also taking a hit from the legal case against them that is being reported on by the media in detail. While Google and Anthropic have executed well in their Generative AI efforts, OpenAI has fumbled their priorities while losing considerable marketshare over the last year. Anthropic’s historical May 6th, 2026 deal with SpaceX AI (formerly xAI) for compute means Anthropic is no longer behind in compute. Anthropic signed an agreement with SpaceX to use all of the compute capacity at their Colossus 1 data center.

OpenAI Talent Exodus Continues

OpenAI keeps losing key talent and this week it was head of private equity Paul Zimmerman, who joined Alphabet Inc.‘s Google as Managing Director and Global Head of Private Equity. Meanwhile, James Dyett, OpenAI’s head of sales, also departed after three years, joining venture capital firm Thrive Capital as Operator in Residence. OpenAI had already lost several key executives and AI researchers earlier in 2026. Meanwhile its poaching more corporate types and executives from SaaS companies in prep for its IPO. I’m fairly skeptical that their Codex campaign on X is even true.

We have to compare OAI with Google, Anthropic and the top Chinese labs at every step now not just in model benchmarks, but in product execution, confidence in their business model, marketshare and revenue momentum (ARR) vs. cash burn. Just months for a supposed IPO later this year, OpenAI do not appear product or confidence ‘ready’.

Anthropic not OpenAIm is the AI Company Everyone is Talking About

While SpaceX, OpenAI and Anthropic are all going public in 2026, Anthropic is the one that is actually performing in the space. explained what Silicon Valley and China are noticing too:

• Anthropic Q1 alone $10B ARR to $30B.

• Anthropic in April, 2026 went from $30B to $44B and that's $96 million in new ARR added every single day.

• OpenAI has not seen similar ARR momentum, in fact their revenue growth has been slowing as they have lost markets to Google and Anthropic in the last year in key areas.

Anthropic on Pace for a $2 Trillion in Revenue by 2030

Read more

Good Morning,

If Generative AI is a general purpose technology (GPT) we should be seeing massive job creation soon. AI is supposedly “hollowing out” of some entry level positions in some industries is while simultaneously revamping many roles within companies. Some areas like Software Engineering and customer service are on the front lines, creating new paradoxes of supply and demand.

This week crypto exchange Coinbase said it was laying off 14% of its staff because of you guessed it, AI.

Read the Message to his Coinbase Team

Tech companies doing AI washing due to less demand for their products isn’t anything new. Executives in silicon Valley are using the same arguments on a repeated basis to please shareholders:

“Second, AI is changing how we work. Over the past year, I’ve watched engineers use AI to ship in days what used to take a team weeks. Non-technical teams are now shipping production code and many of our workflows are being automated. The pace of what's possible with a small, focused team has changed dramatically, and it's accelerating every day.” - Brian Armstrong, CEO, Coinbase.

So where are all the new jobs this technology is supposed to be creating? How much should we be listening to Economists who know how technological shifts occur historically and what to expect in the near future?

Torsten Slok, the chief economist at Apollo Global Management, has been one of the loudest voices in the "AI is additive" camp. He’s been weighing in morelately on the labor impacts of AI. He recently oddly enough compared Chinese manufacturing automation of the early 2000s to Generative AI (which isn’t a good comparison at all) and said:

“The AI shock is following the same playbook. The displacement force is different this time, impacting cognitive and white-collar work rather than factory floors. But every other element of the structure is remarkably familiar: a powerful disruption, immediate job losses in exposed sectors, and a wave of offsetting gains that keep headline unemployment low, see chart below.”

Is the history of Technological automation repeating itself?

But if this is a Jevons paradox type situation how about wage gains and the quality of the new jobs created? More like increased demand for more efficient low cost jobs (like in this other case study he cited). But where are the entirely new jobs going to come from?

Is Jevons Paradox Implicated in demand for jobs made more efficient by AI?

If the Generative AI era began in 2017 but functionally in 2023, we are still very early. The impact is still highly uncertain. If SWEs are more productive now with AI coding, you’d think demand for them would pick up again after the initial dip. If AI is transforming roles within tech companies, wouldn’t you expect they need more Product Managers?

• Paradoxically, as AI makes software development faster and cheaper, the demand for PMs often increases. While demand for process orientated & focused roles in Product Managers has decreased, demand for strategy and creative type PMs has increased.

Read more

Editor’s Note:

In the early 1960s, John F. Kennedy faced a technological anxiety remarkably similar to our own. Back then, the “threat” wasn’t Large Language Models, but the “Cybernation Revolution”—…

Read more

Hey All,

Welcome to another viewing of AI through the lens of infographics. This visual approach lets you take a leisurely viewing of AI’s rapid progress in the last few months. I include my field notes on more contemporary issues as well including curation from multiple sources.

This is part 3 in our series.

• Part I: Summary of the AI Index Report 2026

• Part II: Summary of the AI Index Report 2026, Part II

BigTech Capex Revised up another almost 30% so far in 2026

I’m sort of obsessed with Earnings and Capital Expenditures of the world’s biggest companies who are driving a lot of the AI momentum with their spending and choices and frankly, just trying to meet capacity as it explodes.

Artificial intelligence investment is projected to keep growing at a parabolic pace through 2027, with Wall Street analysts now estimating $1 trillion in total spending, reports CNBC and others.

1 Trillion Capex by 2027 is in View

This is of course like I have mentioned is considerably higher if we include other big players like Oracle, CoreWeave, Crusoe, xAI, Nebius and other emerging Neo Clouds. While BigTech is often the sponsor of these additional capacities it does not account for all by any means. A lot of Oracle’s building out of datacenters relies on its own (considerable) debt and rather speculative third party debt by Softbank. Do note that Crusoe is likely to go public in 2026 and while not as big as CoreWeave is fairly aggressive in its approach.

This acceleration of Capex directly correlates with an increased demand for compute in the age of inference. For every meme product like OpenClaw, that’s a lot more tokens in the “tokenmaxxing” fad of the Generative AI movement. Large amounts of corporate bonds and Venture Capital funding are essentially subsidizing all of this burning of tokens and heating of GPUs. I don’t think anyone has the real numbers.

But what’s clear is that it’s going to snowball and it’s designed to do so to legitimize the Capex where corresponding increases in BigTech Cloud computing revenue and Ads revenue is occuring. The actual ROI for society remains somewhat more uncertain and speculative. What kind of productivity gains can we really expect and how does the ROI diffuse into society and the labor market? A lot of the labor implications of AI is fairly ambivalent, uncertain and mixed. Which leads us to some fairly unhinged Op-Eds recently.

AI Job Loss or Jevons Paradox?

Case study: AI’s impact on customer service seems to suggest an increase in labor demand.

Gen AI May not be creating New Jobs but More of the same

According to Torsten Slok of Apollo, nearly two million workers in the Philippines now work in call centers, up every year since 2016 and through the AI boom. This is Jevons paradox in action: as AI makes call center work cheaper and faster, companies are buying more of it, not less. Lower cost per interaction does not mean fewer interactions. It means more customers served, more channels opened and more markets worth reaching. The technology that was supposed to shrink the industry is fueling its expansion.

• What will be the labor economics of agentic AI? After an initial drop in SWE hiring, demand has started to pick up again for SWEs.

• Demographics and immigration changes in the U.S. labor market appear way more significant than AI’s impact of the last few years. A low hire environment persists.

• Aging populations and reduced immigration could hurt the creation of new jobs in the United States in the 2026-2030 period, irrespective of what AI does toa automate certain tasks. Healthcare remains the primary driver of employment.

However many major Tech corporations are conducing layoffs to compensate for their spending on things like AI Infrastructure. You can read a couple of pieces in the NYT that caught my attention on the future of jobs here and here. The “abundance bros” seem to have gotten a lot of things right and wrong in their thesis.

In essence in the customer success example, more efficiency appears to drive more human and AI interactions due to lower costs. Does Jevons paradox save some of the pain of the lower demand for entry level positions or just displace (and even accelerate) them to other parts of the world?

Is AI Augmenting New Business Formation?

“Sectors with the highest AI adoption rates have also seen the strongest growth in new business applications since 2022, showing that AI is lowering the barriers to starting a company, see chart below.”

• With the GenZ college graduates finding the job market more difficult resort to starting more new AI enabled new businesses? GenZ have already grown up with more entrepreneurial curiosity and the U.S. labor market might be pushing them more in that direction.

• A lot has been made on social media about AI augmenting solopreneurs, though it’s not clear if AI-enabled solopreneurs is a real trend. What do you think?

Leave a comment

What’s going on with Anthropic’s pre-IPO Valuation?

• Recently SemiAnalysis placed Anthropic’s ARR as high as $44 billion. Anthropic is looking to raise another round of funding before its IPO at a valuation of near or around $900 Billion. Given the soaring demand from investors seeking a stake in the company, the final valuation may well exceed that figure and be closer to the private valuation some sources are citing to $1 Trn.

Anthropic’s Mythos Model Opposed by White House 🏛️

According to Anthropic’s own researchers, Mythos is powerful enough to identify and exploit sweeping cybersecurity vulnerabilities, potentially giving hackers a major advantage. While the Department of War feuds with Anthropic, it appears the Trump Administration is against the release of the full model.

The White House's opposition is currently one of the most high-stakes "informal" regulatory battles in tech history. Anthropic allegedly admitted recently that it was investigating a potential unauthorized access to its Mythos model. As far as we know in early May, 2026 the Trump administration officials said they oppose Anthropic’s plan to expand access to Mythos to roughly 70 additional companies and organizations, citing security concerns. However Trump has known ties to Peter Thiel and thus to OpenAI’s Sam Altman. The Pentagon somehow marked Anthropic as a National Security priority supply-chain risk. CNBC and others note it is not clear how the DOD could use Anthropic’s Mythos model without violating the supply chain risk designation.

It’s highly controversial because it’s escalated and some policy analysts are calling this the "Mythos Block." It is the first time the U.S. executive branch has restricted a software product's distribution without a specific law or court order. All of this is highly unusual. Never have we seen National Defense, politics and AI intersecting quite like this before.

This and meanwhile, as the Pentagon has reached a new AI deal with seven (competing but inferior) other companies. They are SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft and Amazon Web Services. These agreements are designed to integrate advanced AI capabilities into the military's most sensitive classified networks (Impact Levels 6 and 7).

Let’s dig a little more into this.

Read more

Good Morning,

What is BigTech going to become exactly?

When I started this publication around four and a half years ago, the term “AI Supremacy” meant to me the competition between the U.S. and China. However the American stock market is pointing in another direction.

What’s becoming more clear is that hyperscalers are the true landowners of AI. They are the custodians and the ones that benefit directly from the demand for compute in more Cloud computing and digital Advertising revenue. This week’s Earnings bonanza of BigTech (4 of the Mag 7 reported after the close yesterday) really displayed this 4 years later (ChatGPT debuted at the end of 2022) in more concentrated terms than ever before. A new world of technological domination is coming into view.

This means essentially that AI is driving unprecedented centralization not just in Silicon Valley but that has taken over Washington and the direction the U.S. is taking on the global state.

4 Years Later (Gen AI snapshots)

• Capex keeps accelerating, like in the case of Meta guiding upwards.

• Cloud Computing revenue growth is meaningfully accelerating

• Digital Ads efficiency is improving a lot due to AI, e.g. at Meta and Amazon

• The demand for compute is accelerating faster than capacity can keep up with

• Datacenter impacts on GDP is concentrated in just a few companies

• BigTech with their capital expenditures to build AI Infrastructure control the pace of the AI movement and the majority of most impactful AI products

• They are also the ones who own major equity stakes in the likes of Anthropic, OpenAI and other AI startups who will become among the most (also concentrated) global winners in the AI arena

• Tokenmaxing and Inference heavy demand is pushing bigger AI datacenter clusters into being

The landowners of AI also include of course the GPU and ASIC makers (AI chips), so Nvidia, Broadcom, TSMC and Google with their TPUs here becomes way more powerful in the 2020s. The rise of the (Philadelphia) Semiconductor Sector ETF displays a Semiconductor boom almost entirely controlled by American incentives. Ticker SOX. It’s up around 50% so far just in the first four months of 2026. Historical actually.

Read more

Hello Everyone,

A little introduction -

• OpenClaw is a free and open-source autonomous artificial intelligence agent that can execute tasks via large language models (LLMs), using messaging platforms …

Read more

Good Morning, everyone

Thank you for supporting this publication. This is my primary income and I do my best to help you stay informed and keep you stimulated about AI at the intersection of business, society, technology and the future. The usual time the Newsletter goes out is 05:30 AM EST. However this can vary, more so in the Spring and Summer months.

I want to continue our visual tour of some of my favorite infographics and updates of the AI Index report, as well as other things related to the U.S. labor market, macro, geopolitical, the economy, future of jobs, GenZ trends, and related interests, primarily in infographic form.

To read the first issue of this series, go here.

You Don’t Say </>

The Federal Reserve just put a number on something developers have been experiencing for two years. A new study by Fed economists Leland D. Crane and Paul E. Soto found that employment growth among U.S. programmers dropped roughly 50% after ChatGPT launched in November 2022. Before that, programming-intensive jobs were growing at around 5% annually—well above the overall labor market.

Why Infographics

The Daily Spark by Apollo by Torsten Slok is another of my favorite sources for economic and macro infographics. What you find is not all Economists are honest in their incentives. The bifurcation of young people, post graduate entry level employment and AI really has me worried. It’s not the concerns you find in VC AI reports or Academic instructions with affiliations with BigTech. That’s part of the story I refuse to not tell. Infographics can quickly help orient ourselves among multiple data points that converge across spectrums of tech, economics and the future of jobs.

Because I want to do my best to display balanced coverage including some contrarian and skeptical insights that are purely my own and aligned with my own conceptual frameworks, this format of curation favors that.

Why so Anxious? AI is causing systemic division

AI can destabilize a consumer based economy, weaken democratic values and continue to warp Neo Capitalism, and it looks problematic for social unrest, wealth inequality and general rule of law. So why is that?

What happens if a kind of AI populism rises against the AI revolution? There’s a significant disconnect between Silicon Valley, Washington and the general population and real citizens, consumers and young people on this. Axios had this paragraph that stopped in me in my tracks: (I’m going to start a bit critical and realistic but I’ll eventually get to the AI Index part).

“Half of Americans are more concerned than excited about the use of AI, according to Pew, and Gen Z’s excitement about the technology sits at 22%, according to Gallup.”

It’s not just an AI paradox, it’s a legitimate AI risk that I’ve not seen Think Tanks and AI policy academics report on much. Bottom up consensus on AI is starting to form, and it’s contrary to what a lot of Executives, elites and BigTech lobbying has been suggesting. It’s this National divide in the U.S. with regards to the benefits and risks of AI that most concerns me.

But Something is not Right…

The more exposed we are to an Internet of AI slop (AIS) and even the potential of career ladder disruption, the less happy young people are starting to be with regards to their trust and hope that AI will bring a better world.

There’s a point where digital well-being plummets and the affordability crisis starts to impact youth unemployment and the way a culture sees its future with the prospect of AI. A Generative AI that’s not creating meaningful jobs, opportunities and is clearly weakening democracy. This isn’t a warped kind of AI populism, this is AI impacting the very human condition it’s supposed to be empowering, enhancing and augmenting.

We know that the Trump Administration created Energy Crisis with its attack on Iran with Israel is going to meaningfully increase inflation.

Eight weeks into the 2026 conflict (Operation Epic Fury), the global economic data suggests that the "disinflation" trend seen in early 2026 has been reversed. This should also worry those young graduates who are struggling to find entry level work in their industry because U.S. Administration policies and possibly even AI are pressing down their opportunities to participate in the labor force.

Why are Americans growing weary of AI?

• How Americans view AI is deteriorating rapidly in 2026. Drawing on five years of Pew Research Center surveys - I believe the low consumer sentiment on the economy is seemingly being paired with the AI boom, in a K-shaped economy.

• As of late 2025, 50% of U.S. adults feel more concerned than excited about AI’s increased use in daily life, while only 10% are primarily excited. As journalism is being eroded and VC-media is on the rise, if you follow LinkedIn or Twitter/X you’d think it’s a lot higher than those 10% who are excited. The American internet is no longer reflecting reality to such an extent GenZ are beginning to leave the legacy apps that started the internet.

• The discontent around AI could signal the fall of the legacy U.S. advertising based internet.

• A rise in AI anxiety stems from multiple issues, including degraded work opportunities. 18% of workers now believe it is "very or somewhat likely" their job will be eliminated by AI within the next five years, up from 15% in mid-2025. Anthropic, OpenAI or Microsoft making dire predictions isn’t helping and feels particularly tone dear and particularly harmful to the actual experience of people.

Got to give Tim Cook some Credit for Apple Management

Steve Jobs was a tough act to follow but Tim Cook did as well as anyone could have expected. Let’s give credit where credit is due though, Tim Cook’s legacy is being cemented as one of the most successful corporate stewardships in history. Essentially, Cook transformed Apple from a "disruptive hardware company" into a resilient, $4 trillion global powerhouse. Take a bow.

Tokens > People & Workers

Read more

Good Morning, everyone

While I’m trying my best to post a bit less with AI news saturation happening, some days do…

Read more

Good Morning,

The buzz around AI coding is reaching a feverish race for AI Supremacy in mid 2026. Just weeks from the onslaught of the biggest AI IPOs in history, the future of vibe-working is shaking the foundations of our anticipation of what becomes possible with this technology.

As much attention as Anthropic is getting, Cursor is the AI startup that I find most intriguing. While Anthropic needs to extend its partnership with Amazon (AWS) for more capacity, Cursor who famously refused to be acquired by OpenAI, is in a unique position.

Cursor (the flagship product of the startup Anysphere) is building one of the most viable “vibe-working” platforms in the world. This is the sort of interface and tech that can enable one high performing employee to do the job of what would have taken entire teams just a few years ago to accomplish.

Download Cursor for Windows

In a shocking turn of events, SpaceX is claiming a deal with Cursor giving it the right to acquire it for $60 Billion. Is that so?

But I doubt very much Anysphere, the team behind Cursor would let itself be acquired. I asked to explain for us one of the key new things Cursor is building. While Cursor has a reputation as one of the best AI coding tools out there, it’s building something far more valuable to various kinds of professionals. I personally think it’s worth a lot more than $60 Billion.

Separately and before this was announced, Cursor (Anysphere) is seeking $2 Billion in more funding at a $50 Bn. valuation. Now this week, SpaceX said it has struck a deal with Cursor to develop a next-generation “coding and knowledge work AI,” which includes a surprising provision — an option to buy the popular software development platform for $60 billion later this year. But it sounds to me like just a model training partnership.

Cursor’s new funding round is likely to be led by Andreessen Horowitz where Nvidia and Thrive Capital are also expected to participate. I would be fairly disappointed if Anysphere was acquired just as it is showing so much momentum. They have youth and talent on their side.

So four MIT students who bet that the "AI extension" model was fundamentally broken could all be Billionaires by the end of the year. Or they could go on to build the future of Enterprise AI.

What do you think?

Leave a comment

What a Crazy Partnership - but it makes total sense for both parties

writes the AI Augmented Engineer.

The AI-Augmented Engineer

Accelerating software engineering careers with AI workflows. I show you how to use AI to write better code, ship faster, and get ahead.

By Jeff Morhous

• Cursor’s recent pivot is just like Codex

• Claude Mythos is genuinely terrifying

• Tokenmaxxing at Meta

• You’re only using 20% of Opus 4.6

View Cursor Docs

Cursor’s Momentum now has SpaceX Funding

SpaceX working with Cursor gives it an edge to catch up to Anthropic in Coding AI, while Anthropic benefits from the API usage and reliance of Cursor users on Claude models. Google Deepmind also created a “strike team” to catch up in Coding AI. Code red for OpenAI’s Codex is good for the future of AI coding which powers the momentum of Agentic AI.

• On April 20th, Alibaba Qwen recently announced Qwen3.6 Max Preview that brings stronger world knowledge and instruction following, along with significant agentic coding improvements across a wide range of benchmarks

As much attention as Claude Code, Cowork, Design and other versions of its product get, Cursor is building a very integrated platform that is quickly getting more usage, but less attention. Anysphere doesn’t have to worry about pre IPO marketing or capacity issues.

Read the Story

Anysphere is Nailing AI Product

Unlike OpenAI, where Cursor really shines is in product innovation. You can feel the difference in talent. Cursor has recently transitioned from being an AI-enhanced code editor to what they describe as an "Agent-First Interface." The pivot is a vibe-working masterpiece.

1. A week ago Cursor announced Interactive Canvases (Cursor 3.1).

2. On April 13th they announced Tiled Layout Interface and improved Voice input.

3. In early April Cursor announced the new Cursor. With Cursor 3, they took that a step further by building this new interface from scratch, centered around agents.

The Best Interface for AI

Cursor Glass is the codename for Cursor version 3.0, a major evolution of the AI code editor released in March 2026

Try to wrap your mind around what this means Cursor is becoming. It’s in some ways better than Codex and Claude Code. It’s the best unified platform for vibe-working and I believe its ARR over the next 18 months will reflect that. Listen to them 1 year ago on Lex.

Anysphere is the one company where the smartest young people in the world are building the future of AI.

Key Features of “Cursor Glass”

• Agents Window: A dedicated space where you can run multiple agents in parallel. For example, one agent can refactor your backend while another builds a new frontend component, each operating in its own execution thread with isolated context.

• Cloud Handoff: This allows agents to switch seamlessly between your local machine and a cloud environment mid-task. A “Cloud Agent” can continue working on complex computations or long-running tasks even after you close your laptop.

• Composer 2: The underlying engine (powering the “Glass” interface) that delivers faster coding performance and better coordination for multi-file edits.

• Design Mode: A visual-to-code feature where you can point at a UI element and describe changes in natural language; the editor then generates the code and provides a live preview for approval.

• Unified Interface: Instead of standard menus, the interface prioritizes a prompt box and an integrated “Mission Control” grid view of all active agent sessions and task threads.

The Big Picture

While Cursor is raising more funding as first reported by Bloomberg it’s now working with SpaceX, it’s already in close partnership with Anthropic. Cursor and Anthropic have historically had a close, collaborative relationship, often described as a form of "co-opetition"—simultaneously being partners/customers and competitors.

While Codex claims that it’s getting a surge in users or customers, OpenAI is in a downward spiral. Anysphere could be the major winner of Enterprise AI, not named Anthropic.

Last week, it was reported that xAI would begin renting computing power from its data centers to Cursor.

Cursor to me is more than just an AI-native code editor by Anysphere in 2026. It’s becoming valuable to professionals outside of just SWEs, like Product Managers, Designers, vibe-coders and agent orchestrators of various kinds including solopreneurs.

According to the deal, If SpaceX chooses not to exercise the buyout (at $60 Bn.), they have committed to paying Cursor $10 billion as a fee for joint development work. Honestly I don’t think SpaceX will be able to afford that acquisition due to how much money xAI is burning each month. Due to how costly orbital compute is going to be. I’m highly skeptical SpaceX can exercise the buyout option even if it wanted to.

This is easily the biggest AI news that happend this week by the way. Simply because Enterprise AI and vibe-working is the biggest AI trend of the mid 2020s.

SpaceX described the partnership as a project combining Cursor’s “product and distribution to expert software engineers” with SpaceX’s Colossus supercomputer…

In my opinion Cursor will grow revenue in 2027 like Anthropic is growing its revenue in 2026, a huge acceleration. It’s insane to give away the company for 60 Bn. It’s possible that OpenAI Startup Fund (Sam Altman) and a16z have pressured them into taking this deal. The only buyer that I would be satisfied with is if Nvidia acquired it.

SpaceX’s founder and CEO, merged the company with his AI startup xAI in February in a deal he valued at $1.25 trillion. But who knows how much the SpaceX IPO will be and likely in June, 2026, just weeks away. xAI is really behind and losing momentum in B2C since their switch to AI coding and a more Enterprise AI approach. Google, OpenAI and China is also going after the same TAM (total addressable market).

Jeff is a senior software engineer but teacher at-heart. He spends his working hours solving hard software problems using AI tools, but spends the rest of his hours teaching other developers how to do the same.

The Deep Dive into Cursor Glass

Jeff goes step by step why the new interface and platform Cursor has built is different. As a SWE and writer/teacher he has a more much nuanced understanding of this than most sources.

Cursor’s partnership with SpaceX is a compute story. Cursor will use Colossus—which utilizes roughly one million H100-equivalent GPUs—to train its next-generation "Composer" models.

With access to this level of compute, will Anysphere become a hub within SpaceX or will it build something else entirely with its own trajectory? We’re at a significant cross-roads in what the future interface of AI even looks like in 2026.

Cursor Glass is by all measures a deviation from all that has come before. Essentially, it marks a shift from a standard IDE with an AI sidebar to an "agent-first" workspace designed for parallel development. This agent first workspace is what I call a vibe-working native platform.

Cursor is the likely tool that makes mainstream Multi-Agent Parallelization. So instead of one AI working sequentially, different agents can now handle specialized tasks (e.g., one on database schema, another on API endpoints) simultaneously. That’s the point where vibe-working makes companies more productive at scale.

There, I actually said it. It doesn’t surprise me that the talent rich team at Anysphere are the ones to build it.

Read more

Good Morning,

I want to do a visual tour of some of my favorite infographics and updates of the AI Index report. I think it would be fair of me to suggest that this year’s report emphasizes that while AI capability is accelerating, the governance and safety frameworks meant to manage it are struggling to keep pace.

The reason this exists is the AI Index came out of a project of the One Hundred Year Study on AI at Stanford to serve as a faster-paced annual lens on advancing AI capabilities.

Browse the Slides

The jagged frontier speaks of a quickly developing industry that is a great lens on the State of AI over the past year 2025-2026 and the big picture. While a lot of the AI reports I read are from Venture Capital funds whose job it is to make AI look amazing, with an academic institution we can hope to find a less biased snapshot. I am not sure if that turns out to be true however. The AI Index that was established in 2019 as an independent project claiming to provide unbiased, rigorous data on the progress of AI. You can follow the Stanford Institute for Human-Centered AI (HAI) here. View their latest news here.

AI News in a Nutshell

Last week Anthropic released:

• Claude Opus 4.7

• Claude Design

What to use Claude Design For?

Initial test use cases for the tool are as follows:

• Realistic prototypes: Designers can turn static mockups into easily-shareable interactive prototypes to gather feedback and user-test, without code review or PRs.

• Product wireframes and mockups: Product Managers can sketch out feature flows and hand them off to Claude Code for implementation, or share them with designers to refine further.

• Design explorations: Designers can quickly create a wide range of directions to explore.

• Pitch decks and presentations: Founders and Account Executives can go from a rough outline to a complete, on-brand deck in minutes, and then export as a PPTX or send to Canva.

• Marketing collateral: Marketers can create landing pages, social media assets, and campaign visuals, then loop in designers to polish.

• Frontier design: Anyone can build code-powered prototypes with voice, video, shaders, 3D and built-in AI.

You can also use Cursor for Designers that integrates with Figma.

Opus 4.7 is built to be a "digital employee" rather than just a chatbot, not the same as the Mythos model they are still studying. In general, Anthropic recommends Opus 4.7 for demanding use cases where you need frontier intelligence — particularly production-ready code, sophisticated AI agents, and complex document creation. In particular I will note that Opus 4.7 seems to have improved a lot in:

Opus 4.7 as the first “Digital Employee” Type Model

• Instruction following.

• Improved multimodal support: like vision.

• Real-world work; like Finance.

• Memory and using file-system based memory.

• Long-running tasks.

Opus 4.7 is a “Doer”

Opus 4.7 handles complex, long-running tasks with rigor and consistency, pays precise attention to instructions, and devises ways to verify its own outputs before reporting back and as such its fine-tuned to be especially good at agentic coding and related tasks.

It’s likely to be one of the more capable global models of the Spring of 2026.

At its best, the AI Index is a distillation of thousands of articles, research papers, and model data, carefully organized and synthesized.

• 12 Takeaways from the Report

• Want to understand the current state of AI? Check out these charts. (MIT)

Which Companies are Leading in AI?

Furthermore Forbes just released their 2026 (8th annual) edition of their “AI 50” read it here. Supposedly, the iteration of the Forbes 2026 AI 50 list highlights a significant shift in the industry, as leading startups demonstrate a transition from experimental technology to sustainable, revenue-generating businesses. Indeed a few on this list will go public in the next 18 months.

A lot of these have indeed seen ARR growth in the last year that’s surprising to say the least and/or seen valuations grow very quickly. I’m going to link to their LinkedIn profiles so you can quickly learn a bit more about them:

1) Legora - legal automation software (Stockholm)
2) Cursor - AI coding platform (Bay Area / NY)
3) Reflection - open source AI models (NY)
4) Physical Intelligence - AI models for robotics (Bay Area)
5) Speak - AI language tutor (Bay Area)
6) Chai Discovery - AI drug discovery (Bay Area)
7) Black Forest Labs - Image and video generation software (Freiburg)
8) World Labs - Spatial AI model developer (Bay Area)
9) Skild AI - AI systems for robotics (Pittsburgh)
10) Sierra - AI agents for customer service (Bay Area)
11) Mistral AI - Open source AI model developer (Paris)
12) Mercor - Data labeling service (Bay Area)
13) Lovable - AI app and website builder (Stockholm)
14) Listen Labs - AI market research tools (Bay Area)
15) Thinking Machines Lab - AI research and products (Bay Area)
16) Genspark - AI tools for knowledge workers (Palo Alto)
17) Decagon - AI agents for customer service (Bay Area)
18) Together AI- AI cloud provider (Bay Area)
19) Suno - Music generation software (Cambridge)
20) Rogo - AI finance tools (NY)
21) Perplexity - AI search engine (Bay Area)
22) OpenEvidence - AI search for doctors (Miami)
23) krea.ai - Image generation software (Bay Area)
24) HeyGen - AI video generation (LA)
25) Harvey - Legal automation software (Bay Area)
26) Fireworks AI - AI app development software (San Mateo)
27) ElevenLabs - Voice generation software (NY)
28) Safe Superintelligence - AI research (Palo Alto)
29) Midjourney - Image generation software (Bay Area)
30) fal - Generative media infrastructure (Bay Area)
31) Cyera - AI data security (NY)
32) Notion - productivity software (Bay Area)
33) Surge AI - Data labeling service (Bay Area)
34) Gamma - AI graphic design tools (Bay Area)
35) Glean - Enterprise search engine (Palo Alto)
36) Cohere - AI model developer (Toronto)
37) Baseten - AI app deployment software (Bay Area)
38) Runway - Video and image editing software (NY)
39) Crusoe - AI data center builder (Denver)
40) Abridge - AI notetaker for doctors (Bay Area)
41) Synthesia - AI avatar and video generator (London)
42) SambaNova - AI chipmaker (San Jose)
43) EliseAI - AI agents for housing and healthcare (NY)
44) Clay - AI go-to-market tools (NY)
45) Applied Intuition - Self-driving software (Sunnyvale)
46) Cognition - AI coding agents (Bay Area)
47) Replit - AI app and website builder (Foster City)
48) OpenAI - AI models and products (Bay Area)
49) Anthropic - AI models and products (Bay Area)
50) Databricks - Data storage and analytics (Bay Area)

My own take on the slides may include text from the slides and my own commentary at times. I might at times also include slides from other sources that may be related which I will credit. I often will add subtitles to the slides themselves to help the reader orient themselves and to make an editorial point that may differ from the original.

When possible I will show the infographic in a wide-view, resolution permitting. To make it a bit easier on the eyes.

As AI advances rapidly, can the systems built around it keep up?

Let’s begin the first iteration of the summary.

The AI Index 2026

INTRODUCTORY SLIDES:

Three Years into the Generative AI Era, a Multitude of Models Perform Exceptionally Well and are Frontier (SOTA, i.e. state of the art)

OpenAI, Anthropic and xAI Funding Skewed the Charts for 2025 in Private Investment

M&A finally picked up. 2026 and 2027 are widely considered IPO years of note.

While U.S. Leads in Closed Models, China leads in Robotic Companies and humanoid robotic products

U.S. Lead in Venture Capital Private Investments in AI is Significant 2013 to 2025:

• Yet China leads in Open-source LLMs and while it has vastly less AI infrastructure, it has more Energy (that’s considered a bottleneck)

• However the above graph is deceptive because of how funding takes place in China. That is, the private investment figures likely understate China’s total AI spending, as government guidance funds have deployed an estimated $184 billion into AI firms between 2000 and 2023.

The AI Index is one of the best annual resources for tracking real progress beyond the headlines.

Which Industry leaders do you trust? 👀

• See the results of a version of this Poll on LinkedIn.

I hope you enjoy the exploration and AI landscape of insights that it provides. 🗺️

Get the Free Trial

Read more

Good Morning,

I envision a world where LLMs are complemented by many types of AI, Quantum computing and a convergence of new kinds of computing architectures. For all the book smarts of LLMs, they currently have little sense for how the real world works and there’s a growing list of “world model” AI startups trying to fill the gap. People won’t trust a one dimensional AI, more on this at the end.

While Generative AI models have their uses, I believe the next decade will fill in some of the physical AI gaps, and allow for things that were previously not possible. Yann LeCun, Fei-Fei Li and others are working towards this end. The likes of Google, Meta, Tencent, Nvidia and others are of course also developing world models.

Robotic brain startups like Physical Intelligence (π) are building general purpose models for physical AI that are basically foundational models for the physical world to help robots learn among other things. Language isn’t enough and models trained on text have inherent limitations. The token based way of predicting the world is just one perspective. A world-model based AI could have advantages over LLMs that understand casualty, and maintain mental models that are more human-like and grounded. This could enable more efficient decision making AIs, to actually make the world a better place inspired by cognitive neuroscience and grounded in empirical experience.

Yann LeCun, , Packy McCormick and many others make a lot of good points about the inherent limitations of LLMs and the merits of World Models, and I think this is important to point out even as Silicon Valley fuels capex that isn’t rational compared to what the future AI will actually be like. Of course that is just my opinion.

My guest contributor today is of Weighty Thoughts Newsletter, a VC on AI and DeepTech. 📚 His book What you Need to know about AI, recently won this year’s American Legacy Award Best Nonfiction . It was also a finalist in the business technology category. Congratulations to James.

Weighty Thoughts ∘˙○˚.•

Weighty Thoughts

VC on AI, deep tech, startups. Former Bridgewater, Google[x], startup founder. Read by top engineers, fund managers, and policymakers.

By James Wang

My Weights Thoughts of the day ⚗️

A little Quantum tangent on our World Models exploration today:

• Nvidia recently announced Nvidia Ising that the company says might be the world’s first open AI models to accelerate the path to useful Quantum computing. It’s exciting because it’s being piloted by the likes of Academia Sinica, Fermi National Accelerator Laboratory, Harvard John A. Paulson School of Engineering and Applied Sciences, Infleqtion (that is now a public company under the ticker ) , IQM Quantum Computers, Lawrence Berkeley National Laboratory’s Advanced Quantum Testbed and the U.K. National Physical Laboratory (NPL).

Read more

As you know watching BigTech has always fascinated me. In 2026 with the “SaaS Apocalypse” and OpenAI, SpaceX (now including xAI) and Anthropic about to go IPO in months, the hyperscaler incumbents are beginning to come under significant pressure. This is a big deal for the future of technology.

On April 9th, 2026 Amazon’s CEO Andy Jassy, wrote a very interesting shareholder letter. I’m not the type ot typically pause over shareholder letters or earning calls. By the time you read this article you will understand why I think it’s significant. It tells you more about the future of AI than most realize. I think it’s worth analyzing and today I’m going to try to do that. How under threat are the BigTech hyperscalers? Clearly, some more than others.

I really enjoyed digging into the tidbits about Amazon’s evolution in the AI era. If anything it drives home the point of what an “inflection point” these companies believe the Generative AI era represents. They are making calculations that impact many moving parts of their businesses in an evolving competitive landscape.

I go into quite a lot of details trying to model how new aspects of Amazon’s growing business models engages with surging demand (demand for connectivity, infra, model training and more). Whatever is the case, Amazon is pivoting and I consider this capex surge comes at a time of great opportunity for the company that is well positioned in multiple areas with products that directly challenge SpaceX, Nvidia and Google.

The market has had nearly a week to digest the shareholder letter (5 days ago), and Amazon’s stock is up over 13% during this time, hovering around a $2.6 Trillion market cap. In this article I give a sweeping analysis of my key takeaways from the shareholder letter and my own thoughts on what this means for the future. I try to leave no stone unturned. The year of 2026 is a big year for what Amazon has planned. I’m by no means an expert on Amazon but I acknowledges that the drivers of the Generative AI era are potentially a generationally unique window of opportunity for them.

Since Amazon is such a big employer of American workers and such a big part of the lives of American consumers, its rising Capex is as concerning as it is impressive. This article tries to uncover how/if this $200 Billion capex is justified and the Who, What, Where, When, Why it all means. After this read I hope to give you a better 360 degree understanding of what Amazon is trying to accomplish.

Read more