James Thomason

Venture capital has disrupted every industry except itself. My new book is an attempt to fix that. Venture Capital Without Unicorns diagnoses why the model fails most founders, most LPs, and most of the economy, then lays out a specific structural redesign that plugs into today's ecosystem. The contract, the simulators, and the accounting work are all open source. Out now: a.co/d/00hpf7ss

Why being a great analyst does not mean you will be a great investor. Brett Caughran (@FundamentEdge) explains: "There's a key distinction in being a great analyst and being a great investor." "Being a great analyst is a bit more of a scientific process — building models, doing research, meeting management. That's a much more teachable skill." "Judgment and being a good investor is much more difficult. It's much more difficult to evaluate." "It's hard for even allocators to evaluate — is this alpha you've captured for the last three years just a lucky bet? Can you generate that alpha over a cycle?" "Great investment firms have inescapably a marriage of good process and good judgment." "Talent is revealed at investment firms over the years, more than trained." "A typical 30-person fund — maybe 25 are analysts. They're doing the work, covering companies, building models, feeding that process up to the decision makers." "The ones who rise through the ranks are the ones who have that real talent to take investment process and translate it into investment results. A lot of that is just the revealed intellectual wiring of being a good investor over time."

Really sharp piece. The one place I'd push: the 140x ceiling on OpenAI is a signal that equity has stopped doing its job. The reason a 140x doesn't return the fund is the same reason 70% of your seed portfolio goes to zero. Pure equity captures value only at exits, and exits are compressing from both ends. Better picking helps at the margin. But a better contract changes the distribution. There's a third path between concentrating bets and hunting the unicorns. Change the instrument so the middle of your portfolio stops being dead weight. The entire VC ecosystem is downstream of pre-seed and seed investors. Who says we have to play their game on their terms? I wrote a whole book on this: Venture Capital Without Unicorns. Amazon: a.co/d/00hpf7ss

This would make the HPE Autonomy acquisition look smart in hindsight

@SpaceX @cursor_ai This would make the HPE Autonomy acquisition look smart

SpaceXAI and @cursor_ai are now working closely together to create the world’s best coding and knowledge work AI. The combination of Cursor’s leading product and distribution to expert software engineers with SpaceX’s million H100 equivalent Colossus training supercomputer will allow us to build the world’s most useful models. Cursor has also given SpaceX the right to acquire Cursor later this year for $60 billion or pay $10 billion for our work together.

This a really weird way to say you couldn't afford to buy Cursor

@SpaceX @cursor_ai So basically you couldn't afford to buy Cursor.

Here's my reward for backporting my AI pipeline to Gemini. Thanks @Google !

@GaryMarcus How does one boycott the game theoretically dominant strategy?: arxiv.org/pdf/2603.20617

Violence is not the answer. Boycott is the answer. This one’s easy. There is ample evidence that Altman is a dishonest person with inordinate power that we should not trust. But the way forward to is get the board to remove him, not to throw bombs at his house. Humanity must take the high road. The way to take the high road is to stop using his products, in protest of his implied openness to mass surveillance, his mass IP theft, and his company’s opposition to liability for their actions. When people stop using ChatGPT, Altman will have to go; it’s simple as that. Quite possibly the COO or CFO will step in, and we (and OpenAI itself) will all be better off.

@StockSavvyShay @FuturumEquities A PEG of 2 is not expensive nor is a PEG of 0.5 cheap. A PEG above 2 is not a “danger zone”. A company can have a PEG > 2 or < 1 forever and ever.

PETER LYNCH’S FAVORITE METRIC IS THE PEG RATIO PEG < 1 usually means mispriced growth PEG > 2 starts to push into the danger zone Popular software names right now: • $NET ~4.7x • $FIG ~2.9x • $CRWD ~2.7x • $PANW ~2.5x • $PLTR ~2.4x • $DDOG ~2.3x • $SHOP ~1.9x • $SNOW ~1.9x • $MDB ~1.8x • $ZS ~1.6x • $RBRK ~1.5x • $APP ~1.0x • $PATH ~1.0x • $NOW ~1.0x • $CRM ~0.8x • $ADBE ~0.8x • $HUBS ~0.7x • $MNDY ~0.7x • $TEAM ~0.6x • $ZETA ~0.5x

@Rothmus Know who else you shouldnt work for?

@StockSavvyShay @fiscal_ai I don’t think anyone, including Anthropic, has figured out how to do agentic at scale for enterprises. There’s a good chance NOW, CRM, and other saas with data moats figures it out.

$NOW is trading at under 15x EV/FCF while still putting up a Rule-of-40 score of 56 🤯

@simple_traderz @StockSavvyShay @fiscal_ai Memory and chips are late cycle, as is AI infrastructure and optics, though I can see optics running through 2Q. Space has some real names if you choose wisely. IMO, drone defense ftw.

@StockSavvyShay @fiscal_ai I am staying away from SaaS . Current theme is memory , chips , AI infra , optics , space that will be next big

@StockSavvyShay There’s a really good reason $ONDS won’t break $20. Hope that isnt your pick. 😬

Wild to see my $AEHR position up 4x and become the biggest holding in the growth portfolio. I also started a brand new position today. Any guesses? Hint: it’s in the drone thematic.

It's actually time to remove the President from office when he sits down, writes about "ending an entire civilization", and then clicks "send"

@FundamentEdge This is such a good take. The models get you ramped to consensus and key drivers in record time. Your alpha is outside the building. The models free you to do the work that actually results in a differentiated view.

The question of whether AI will generate alpha in the investment process to me is rather obvious. AI can help us: 1) know companies better 2) track them more closely 3) make better decisions Used effectively, agentic AI is clearly a powerful tool for alpha generation. But it's the system around AI that matters. AI doesn't know the answers to my questions, but can enable a process around building much more comprehensive insight in the same block of time. Don't outsource, augment. Like any tool, however, how you use the tool is what matters. I'll give an example. I like to time-block parts of the investment process in my pedagogical approach. How investors spend their time is a critical decision when we all are working with the same ~3,000 hours in a year. I could map out a 300 hour due diligence process on a name, but then I only get 10 shots on goal in a year. If you've been in the seat of a public investor, you know this is a never ending game of triage, particularly as coverage broadens & idea velocity demands increase. When thinking about time blocking your investment process with AI, I would encourage you to do a "before & after" approach. I hear from investors often "I did the AI augmented approach, but I don't feel like I know the name as well". I sympathize with that, and agree, but when I push back the obvious point is that they tried to compress 60 hours down to 20 hours. There certainly will be comprehension lost in translation trying to do that. My suggestion is to time block that entire 60 hours. You are saving ~40 hours, how will you redeploy that 40 hours? Your portfolio won't magically include 3x the number of ideas, so how do you increase the research quality of the existing idea flow? This is where I think top decile investment process will radically change in the next 9-18 months. AI will raise the floor in investment process quality, democratizing "institutional like" quality, but raising the ceiling is what matters: alpha is an adaptive game where power law prevails. Running the same desktop research game with small snippets of primary research won't continue to work, in my opinion. Raising the ceiling, to me, is not about working fewer hours, it's about driving much more comprehensive understanding of the business & the stock in that same 60 hours. Some of this is "what's old is new again" (i.e. for example, multi-manager investment professionals are less present at trade shows than Tiger investors, for example, due to burden of desktop research demands in coverage model). But some of it is thoughtfully deploying brand new AI-augmented processes that can drive deeper comprehension, closer KPI tracking and help drive better, less emotional decision making. To simplify, a compression of desktop research opens up meaningful capacity for more primary research. The critical question then becomes , even if the compression of 60 to 20 leads to some degradation of name comprehension, is that degradation more than offset by the brand new 40 hours of AI augmented primary research. My belief is yes.

@GergelyOrosz Are we really going to ignore the irony Anthropic stole the entirety of humanity’s IP but somehow thinks *their* IP deserves protection?

This is either brilliant or scary: Anthropic accidentally leaked the TS source code of Claude Code (which is closed source). Repos sharing the source are taken down with DMCA. BUT this repo rewrote the code using Python, and so it violates no copyright & cannot be taken down!

Ok, Perplexity Computer is winning. @AravSrinivas how can I help you? Hire me to do something. I need shares.

@morqca @Merridew__ What software is this?

@Merridew__ Not wrong, energy themes are so hot right now

@nickshirleyy I lost IQ points watching the first two minutes of this.

🚨 Here is the full 40 minutes of my crew and I exposing California fraud, Minnesota was big but California is even bigger... We uncovered over $170,000,000 in fraud as these fraudsters live in luxury with no consequences. Like it and share it, the fraud must STOP. We ALL work way too hard and pay too much in taxes for this to be happening. These fraudsters have been able to defraud American taxpayers for years without any pushback from the public and politicians. It is time to EXPOSE IT ALL and end America's fraud crisis.