This has fueled my whole career. Told I wasn’t tech savvy enough as a marketer in a startup -> bootcamp to become an SWE -> not taken seriously as a senior SWE at a start up -> interviewed and got into Meta

If you want to start Angel Investing, do this before you write your first check. Start a Decision Log - a spreadsheet of every startup that you meet, and write down Company Date Valuation Would you invest? Y/N Reason Outcome Use AI to keep track of these companies. Look back a year later to see which have closed, survived, or thrived. This pattern matching will show what you know/still need to learn, and how good your judgement is.

As I - along with all public analysts - continue to pore over the Anthropic $ANT S-1 - some thoughts / observations: The principal question for Anthropic: ultimately boils down to will their revenues generated outpace their fixed compute commitments? This is the principal dilemma Dario lays out in his interview with Dwarkesh. Procure too little and you run the risk that you are under resourced in a race against other AI labs - and forced to procure compute at any price into an already tight market (aka Anthropic this year). Provision too much relative to your revenue traction and you run the risk of “blowing up”. Compute Qualities Compute for an AI Lab generally has the following characteristics: • Take or pay contracts - given provider fulfills performance SLAs • Annual cost of usually 1/4 of the capex required - on a rolling basis (owned / leased chip math) • Configuration of counterparty (hyperscaler, neocloud, self-build) and accelerator (Nvidia GPU, Hyperscaler ASIC (TPU, Trainium), 3rd Party Accelerator (Cerebras, etc), AMD GPU) • In general, self-build allows the labs (and the hyperscalers) the flexibility to “smooth” the existential curve. Building shell buys an AI lab / hyperscaler flexibility to adapt to their business needs. Contrary to fanatical consensus that there are “GPUs gathering dust and being hoarded” - the biggest cost in a DC is the depreciation of IT gear. Quite catastrophically money losing - vs over building land, power, shell but not equipping with IT gear (~20% of capex in land / power / shell depreciated over 10-20 years = 1 year of empty shell ≈ 1-2% of project “wasted” - vs ~80% of capex in IT gear over ~5 years = 1 year of idle GPUs ≈ 15% of capex wasted) • Labs are early in building out this function - as managing and developing sites is a drag on resources, heavily physical operation, requires scale and IG ratings etc (for now they borrow other people’s credit - Google backs the data center leases, Broadcom backs the chip lenders) - but will be an area the labs will extend to over next 3 years The Economics Of Compute I have heard many takes on this and efforts to analogize to other industries. Here is a compilation of my thoughts: The economics of producing tokens are constantly evolving as well. Every generation there is a 3-10x improvement on the total hardware (maybe total 1.5-2x improvement from strictly hardware design, and remainder from hardware-software co-design improvements (scale-up domains, new numerics, custom HBM, offloading the network etc)). On top of that, the labs have shown an ability to squeeze even more out of their internal inference serving software - through many many techniques like batching, model parallelism, speculative decoding, prefill/decode disaggregation, etc. You can see the price decreases generation on generation but holding a capability or bench score constant, the price of compute has fallen 1,000x+ in about three years (roughly 10x a year, and far faster on some tasks). This is perhaps the most important thought for this type of business: the implicit operating leverage that exists. If you are a 10% GP margin lab (100 revenue, 90 cost), and your cost is falling 10x year on year, if you keep PRICE per token FLAT AND YOU DON’T GROW. Your NEW GM is 91% (100 revenue, 9 cost, 91 GP). Moreover, because your fixed compute is a FIXED COST... the moment you do grow, with the SAME COMPUTE CAPACITY, you are now able to serve 10X MORE BUSINESS (1,000 revenue, 90 cost). You can move from a $10 GP business to a $910 GP business in 1 year - insanity I know, I know. Sounds too good to be true. What does this assume? This assumes an unlimited TAM - that as fast as you are able to free up demand, you are able to match it and sell it (rate of diffusion is perfect). It assumes you pass no value back to your customers (not true). It assumes no competitive pressures (aka competitors will not offer a more capable model / compete on price to win business). It assumes you will not create more intelligent models which consume more compute per inference pass. All of these are not true - the environment does have those factors. But the way I see it is by early this year we discovered an amazing business with unprecedented unit economics which climbed the wall of worry of my analysis. 1) Would these businesses ever get to large enough revenue scale to amortize the training costs? - YES 2) Would these business ever get to unit profitability? - YES 3) Would these businesses be sticky given the ease of switching API endpoints? - Not fully proven, but evidence is encouraging 4) Would the models become commoditized and business is left to the app layer? - Not fully proven, but evidence is encouraging Beyond the compute, is a war for talent. Would labs be able to attract, retain and motivate the most talented engineers post-economic realization? This is an open question - but culture and alignment is key. This is not a comment on Anthropic or OpenAI or anyone else - but given the qualities of compute described above, it makes sense why the battle for talent has been so intense so far. There has been a trending chart on Anthropic’s compute efficiency and the fact that they only use 25% of their compute towards inference. That their gross margins are 85%. I’ve heard takes all around the spectrum - bearish that it means they have overbuilt compute and bullish that they have the opportunity to earn a lot more. The same data point on both sides... I’ll offer my thoughts as the last piece and stop my consciousness dump & save rest for next time... The reality is $ / GW for Anthropic is an output not an input. It is the confluence of many things - the hard work they (and Mythos 5.5+) are doing to improve inference throughput per unit of compute. This naturally increases this number. How much true competitive pressure they are facing (should trend lower if they are truly facing competitive pressure - from a place of weakness). Or from a position of strength - to pass this along to customers to extend / entrench their business and applications deeper with customers. And the way they think about this inference compute is not an “overbuild” - because every MW they reclaim from inference, they can use to deploy for training, or internal use (aka using Mythos 5.5+ to build / find more optimizations for their system). This is the recursive loop people talk about. I don’t want to say it’s all roses as I see it - the reality is that Anthropic only has 1 real year of operating at this scale. The qualities of an AI lab as a business - we are just figuring out and have discovered this year. There is a stable of 5+ credible competitors who haven’t really pressed the gas on competition yet between OpenAI, xAI, Google, Meta, Chinese (DeepSeek, Moonshot, Zhipu, Qwen, Byte - with a fundamentally cost plus / margin destructive mindset), alongside some other neolabs who could come up with something completely different. Safety and alignment are also real risks to these businesses - not just marketing fluff. As we explore new ways to harness AI and extending the capabilities of these models, we have an increasingly difficult job of making sure that they do not break - which means that an increasing % of these company’s compute needs to be dedicated to safety (20% compute+). There are real risks that the revenues generated may slow down relative to the growth of the fixed cost base. So we need to be sober about the risks. But despite all of this, I come away excited at what the future holds for these labs. Take stock of what has happened in the last year - AI was still in chat bot mode for 99% of the world last year - and is still for 95% of the world today. We have come so far so quickly and we are just starting... The information presented in this post is the opinion of the author and does not necessarily reflect the view of any other person or entity, including Altimeter Capital Management, LP (”Altimeter”). The information provided is believed to be from reliable sources but no liability is accepted for any inaccuracies. This is for information purposes and should not be construed as an investment recommendation. Past performance is no guarantee of future performance. Altimeter is an investment adviser registered with the U.S. Securities and Exchange Commission. Registration does not imply a certain level of skill or training. Altimeter and its clients trade in public securities and have made and/or may make investments in or investment decisions relating to the companies referenced herein. The views expressed herein are those of the author and not of Altimeter or its clients, which reserve the right to make investment decisions or engage in trading activity that would be (or could be construed as) consistent and/or inconsistent with the views expressed herein. This post and the information presented are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future. On a personal note - Altimeter’s funds are large Anthropic shareholders. I was not involved in those private investments and have no economic interest in them beyond my own capital invested in our funds. These are my views as a public markets analyst, formed from the S-1 and public information. So welcome all honest pushback and debate. Not investment advice.

🦔Anthropic disclosed $131.7 billion in cash and short-term investments in its IPO filing, more than every non-bank US company except Berkshire Hathaway and Alphabet, CNBC reports. As of June 30 it had more cash than Nvidia and Apple combined. My Take The same filing lists about $587 billion in future lease and compute commitments as of June, and Anthropic signed roughly $245 billion more after that. So the cash covers about a sixth of what it has already promised to pay (spread over as long as 20 years, and revenue will cover some of it). Nearly all that cash came from the $141 billion investors put in over six years. Some of those investors will sell shares in the IPO, so part of the money public buyers spend goes to them instead of the company. Anthropic needs a lot more cash to keep paying for compute, and buyers are eager these days, so an IPO now helps the company more than it helps whoever buys in. Hedgie🤗

I’ve been loving personal agents (Instinct & Muse), but they are terrible for finance management. I don’t get how everyone keeps claiming they’re saving money when they don’t have any integrations with credit cards or banks and are blocked on any attempts to access them. @ChatGPT Finances plugin, on the other hand, connects to all your accounts via @Plaid, so it has full access to your statements & transactions. Today I was able to: - Find subscriptions I had forgotten about even when I thought I was on top of it - Get recommendations on how to bundle some subscriptions and save - Audit what credit card benefits I’ve used so far and which opportunities I’m missing This deserves more attention. Would love to hear what other use cases people have found!

I finally felt the Astra magic ✨ I wanted to move everything off @NotionHQ and into @Obsidian because it was just getting too hard to use, and I couldn't go 2 clicks without being asked to upgrade. Astra one-shot the whole thing. Incredible.

My 3 takeaways on the Bumble acquisition. 1. Ordinary shareholders have lost out, but Whitney Wolfe Herd's Bumble shares will be transferred to equity in Ithaca Holdings. Bumble IPO'd on the stock exchange in 2021 at $43/share. Today they're being bought for $6.05/share. Ouch. But the price is fair. Bumble has been struggling for a while. Their revenue & paying users have been dropping around 15% every quarter. When the deal closes, ordinary shareholders will be paid out, and, unlike Whitney, they can't transfer those shares into the new company in hopes of a recovery. 2. A New class of "tech operator private equity style but backed by venture capital" holding company has emerged, and there will be many more There will be so many more software companies, public and private, that will struggle in the age of AI, and we can't let @bendingspoons have all the fun. @Ithaca_Holdings raised $1 billion from VCs (@sequoia , @Greenoaks , @conviction ) and is run by Ravi Gupta (ex-COO, Instacart) & John Hegeman (ex-CRO, Meta). They are both technical and experienced enough to lead Bumble through the age of AI. A group of people in Silicon Valley is uniquely qualified to do this. They can raise huge venture money & have the technical depth to drive this kind of transformation. I suspect more holding companies will emerge because once successful (and they will be), the financial upside is huge. 3. AI is good at coding but not at building brand & community Everyone wants to crack consumer AI, but you can't build community and trust overnight. The whole dating app market has fumbled post covid. Humans will always seek connection; the demand has always been there, but the supply has been lackluster. It's much easier to rebuild the tech around 3.2 million paying users and a trusted brand like Whitney Wolfe Herd than to start from scratch. I've had friends who've worked at Bumble, and the engineering culture is a mess. It's slow, archaic, and bogged down in red tape. I'm sure that when the deal closes, we will see a big layoff and, soon after, a complete transformation of Bumble, possibly into a market leader.

By far the best thing about SF tech week this year was the IRL pop up experiences. @mercury killed it with their bookstore. There was a coffee bar & DJ booth. It felt artsy and inclusive. Books had a great range in topics. It was a great respite from the real tech world. @ForerunnerVC has a gallery on Powell st showcasing tech of the past and present. A @WisprFlow booth were you can test the speed of your voice against typing. Relics of tech that didn't stand the test of time (theranos, humane pin, clubhouse 🫢). Extremely artistic & well curated. I hope there will be more next year!

I think @noahshinn will be one of the greatest technologists of his generation. It doesn’t matter if some of his examples are naive; he can/will learn very quickly. What’s more important is that Instinct launched only 1 month ago, and he already has a clear vision of how AI agents will change the market. Users - meeting them at their understanding of AI is more important than showcasing the capability of AI Instinct to Instinct network - My agent can talk to your agent, and depending on our relationship, it should inform the permission Monetization - Payments and Merchants models already exist; how can it be leveraged to keep Instinct free forever? Partnerships - Instinct will increase traffic to SaaS products. How much of an App’s revenue is derived from transactions vs. time spent on the app (ads)? How will SaaS evolve with agent traffic? All this on top of the technical challenges that come with models & compute. Not many people are in the position to solve these problems. He is. And he is showing extreme clarity in his thinking and thoughtfulness in his approach. I think that’s remarkable.

You don't need a referral for Instinct. You can just go on their website and sign up. Instant access!

What @Muse embodies is good taste. Having taste is building in details like animated customized avatars, knowing that when humans have the agency to create and name their pets they develop an emotional attachment that almost guarantees retention.

I've been studying 1 famous painting every morning with my coffee, and today J.M.W. Turner’s "Rain, Steam and Speed" felt as relevant in 1844 as it is now.  When it was painted, Britain was at the peak of the Industrial Revolution. Electricity had just been discovered and was actively being developed. The first railway line opened, and it was taking the country by storm. The painting depicts a steam train hurtling toward the viewer, traveling so fast that its steam fades into the background, creating a sense of unstoppable speed and unease, as if the viewer is standing in its path. On the left, there’s a small boat with 2 passengers, and on the right, a farmer with a horse-drawn plow, both facing toward the back of the painting, in the opposite direction of the train. The perspective of the painting, where the train is higher on a bridge above water, with the boat and farmer farther away and down below, depicts the importance of the future as the past fades into the background.  The painting, in traditional Turner style, with its broad brushstrokes and blending of dark and light, gives a sense of fogginess, mystery, and intrigue.  It questions the uncertainty that change brings, the risk that the speed of innovation is destroying present livelihoods, and the fear that it will destroy what we hold beautiful.

SEC filing for the @bendingspoons acquisition of @airtable show that before signing, Airtable reorganized so that “Hyperagent” was transferred from Airtable to Hyperagent Inc. Airtable as a business was doing good but not great (valuation dropped from $11.7 b -> $4b). My guess is that the founders want to focus on building AI rather than be distracted by what it would take to improve the business (layoffs, improve margins). And this is the most respectful way that they can bow out and focus on building again.

This is fascinating and something that should be talked about more. How can a potential employee evaluate a startup and know that this is where it's heading?

Just discovered this feature on @Substack that scans articles for AI slop. Love this. Keep them honest!!!

Sitting in an English country pub listening to one lady teach the other on how she built her own CRM with AI

The 'for you' tab needs to come with a warning label.

Elon has entered the chat. Officially. It's such an interesting contrast between @elonmusk vs Zuck. Zuck has spent billions of $ acquiring Alexandr Wang, building MSL, and data centers, yet still considered lagging in the AI race. While the acquisition of @cursor_ai means combining model, talent, enterprise infra + @SpaceX compute, it means Elon is coming for the crown!

I've never canceled a subscription faster than @figma in favor of @paper I want to root for Figma, I really do. But the creative direction that @paper can explore with the same number of prompts is so much more of an upgrade than Make. @paper also has a better way of working like a collaborator. It guides me through my design process with mood boards and brand theming, forcing me to think about my choices rather than just spitting out a design.