Steven Xi

@dittycheria Such a shame for Argentina team! I rooted for them (Messi to be exact), and they have come a long way to be in the final. Now I feel they don't deserve it (except Messi). The final would be a better game if England played Spain though.

I was at both the England match and today’s final. It was hard not to notice that Argentina consistently pushes the boundaries of gamesmanship, creating a hostile, intimidating atmosphere for both opponents and officials. It’s a shame, because they have an abundance of talent and hardly need those tactics, yet they seem to be part of the team’s identity. Today, though, Spain refused to be intimidated and was clearly the better side.

@JvNixon Good analysis, Jeremy!

The argument that the US is still ahead: While China has been great on kernel level architectural creativity (GLM 5.2 IndexShare, Kimi Delta Attention, Qwen's Gated Delta Unit, Deepseek's Sparse Attention), OpenAI, Deepmind and Anthropic have been great at more conceptual, higher level research like reasoning models, RLHF and MOE. Kimi is heavily distilled from Claude and at no point in history have the Chinese had the strongest AI model (or even been in the top 2) available to the general public. Anthropic has models trained that are past Mythos in quality that are unreleased (through which foreign employees have been able to continue using SOTA models despite government bans of specific model names).

Well said!

Why don't we have the US Department of Recruiting that recruits the best engineers in the world to America? This has never made sense to me. Do we want to win or not?

@chrisyeh I got this phishing text this morning too! Fortunately, I know your # and the text came in from a strange #.

Two of my friends received phishing attempts impersonating me. They received texts saying, "Hey <name> How's everything going? It's Chris Yeh". Fortunately, they could tell it wasn't me from the lack of punctuation. I would never end a sentence without a punctuation mark!

@VoidAsuka @MiniMax_AI Incredible woman. Hope she recovers soon!

let me introduce the toughest woman in open-source ai: the lady behind @MiniMax_AI literally went to the hospital for the launch of the m3 model. 😭 never look down on Chinese women. they might be as pretty as an influencer, or they may look small and plain - either way, they are making big contributions to the tech scene.

Folks, let us get ready and make it happen!

@bernhardsson Plenty talents in Europe for sure and Swedish are leading it! Can someone tell me why? @MaxJunestrand

100%. I want to short every startup that is trying to be “x for Europe”. Just build a better x!

Always love @moinnadeem's clarity! What do you think?

@moinnadeem Always love your clarity, Morin!

Modal's advantage over all of these companies is bundling. I never even looked at the other agent sandbox providers: we run Phonic's inference workloads on Modal, I have our sales pipelines on Modal, agents were just a natural fit too. Why use 3 providers when 1 suffices?

@ttunguz Congrats, Tomasz! Let us find a time to celebrate it and catch up too!

Three years ago, we launched Theory Ventures with a simple premise : AI would reshape how software is built, sold, deployed, & operated. Within that world, we would build a concentrated, thesis-driven firm. The market moved faster than even the most bullish expectations after the ChatGPT moment. Frontier models leapt from delicate demos to production systems. Open source models have become substitutes for enterprise workloads. Inference emerged as the dominant market in AI. Underpinning all of this, AI compresses time. New models are released every 41 days. Companies reach $100m in revenue in record time. We all achieve more faster. In celebration of our anniversary, we wanted to trace that mechanism through the market shifts of the last three years. The first casualty of compressed time is the old language of venture capital. Seed, Series A, Series B categories still exist, but they describe the financial product companies seek rather than rather than company maturity. Venture firms have left the idea of offering a standard financial product to bespoke offerings : seeds range from $1m to $500m in size. Can we really call it all the same thing, anymore? Three years ago, a seed company was often a small team with a product concept & early signs of product-market fit. Today, some seed rounds are larger than IPOs, fueled by great ambition, a supportive VC ecosystem, & the promise of generational scale businesses to be built. Part of this is inflation in private markets. But more of it is time compression : the best companies mature much earlier than software companies did in prior generations. We’ve learned as an ecosystem how to build software companies & AI accelerates product development. Compressed time also redraws the map of where great opportunity lies. When we first launched Theory, most AI conversations centered on models. Remember the debate of whether model companies would be the airlines of the era? Today, inference is becoming the dominant market. The market is segmenting because the workloads & buyer preferences have evolved - very few companies can afford state-of-the-art AI for everyone - & each specialized constraint creates a new infrastructure category. Companies like @sailresearchco are building the systems that operationalize intelligence : serving it cheaply, routing it intelligently, & specializing it around use cases like video, batch, local, agentic, & real-time workloads. Databases followed this path a decade ago. They fragmented into OLTP, OLAP, vector databases, & streaming systems. Those markets have evolved with AI, a pattern we’ve backed through @motherduck & @lancedb , with @omni in the AI analytics layer above them. Inference infrastructure is now specializing the same way. The expense of inference reinvigorates a sedate market that has been controlled by behemoths for a decade : advertising. Every major interface shift, TV, web, mobile, streaming, found its answer to monetizing a massive audience in ads, & AI is no different. AI advertising is emerging as the subsidy for inference costs, letting applications grow usage & revenue together rather than against each other. We wrote about this dynamic when we led @koahlabs ' Series A : native ad formats inside AI conversations are producing click-through rates 4-5x the display baseline, & an agentic app builder can provide inference offset by ads. The same compression closed the gap between closed & open models, cloud models & local models. The conventional narrative holds that frontier closed-source models lead & open source follows. We’ve reached the iPhone 15 moment of AI. Many models are good enough for most work. Running a model locally reduces cost, improves latency, increases control, & minimizes data governance concerns. Enterprises are adopting local & open-source models for sensitive workloads, & frontier capabilities compress toward consumer hardware within a few years. What once required a hyperscaler cluster runs on a laptop just a few quarters later, a shift @ollama brings to millions of developers. The promise of AI is that software will ultimately be more secure : machines that read every line of code, patch faster than attackers move, & never tire. In the meantime, the attack surface is exploding. MCP servers, skills, plug-ins, & coding agents each introduce new entry points, & enterprises are deploying them faster than security teams can review them. Attackers are massively parallel & shrinking necessary response times from months to minutes. Defenses must respond. It’s why we backed @DropzoneAI , whose AI analysts investigate the alert flood no human SOC can keep up with, @Maze_Security , which applies agents to cloud vulnerability triage, & @artemis , securing the new agentic surface itself. The same agentic wave is rewriting operations. ERP & back-office systems have resisted change for decades because the work is unglamorous, the data is messy, & the switching costs are enormous. One CFO we interviewed, when asked about a startup said, “that company has only been around 15 years; they are too immature.” Agents invert that math. Systems that read documents, reconcile records, & execute workflows can attack operations from the inside rather than demanding a rip-&-replace. It’s the thesis behind Doss, rebuilding ERP for teams that move at modern speed, & Backops, applying agents to the back-office work no one wants to do by hand. AI has impacted crypto, another market fueled by data. Prediction markets, stablecoins, micropayments all have an AI infusion to them. Today, crypto companies need to generate revenue & use AI to provide better experiences, which led to our investment @AlliumLabs , the data layer underneath that institutional wave. Recognizing shifts early requires fingers on keyboards, wrestling AI agents into compliance rather than observing it. We built Theory as a technical organization, experimenting with AI across research, sourcing, diligence, portfolio support, & internal operations. Working inside these systems sharpens our understanding of where the stack is breaking & where new workflows are emerging, while deepening our empathy for founders deploying real AI systems inside enterprises. It’s harder than social media says. AI also changes the economics of an investment firm. Over the last decade, venture firms scaled by adding people. AI-native companies are demonstrating that much smaller teams can operate at 10x+ the leverage of prior software generations, & the same dynamic applies to us : since launch, we’ve analyzed 2x the investment opportunities with a team of just 3 investors working alongside a nine-person intelligence organization. None of this works without the team behind it. Theory started three years ago as a handful of people & a thesis. Today we are thirteen strong. We believe this is the structure of a modern venture capital firm : engineers & researchers who build the systems we use every day : agents that map markets, pipelines that surface companies months before they raise, & research infrastructure that lets a small team cover the ground of a firm several times our size. Everyone at @Theoryvc works with the technology we invest in, & that shared fluency shapes every decision we make. The firm we’ve built over three years is itself a product of the thesis : a small team, deeply technical, operating with the leverage AI makes possible. But the real story of these three years is the founders. They compressed decades of company-building into quarters & shipped products that rewrote what enterprises expect from software. The next three years will make these look slow. The most ambitious builders we meet are just getting started, & we can’t wait to see what they do.

If you are an AI researcher, come and join us this Saturday afternoon! As a sponsor, we @eastlinkcap are happy to talk to you!

data is all you need. our back of napkin math on the market: ultimately 4-5 multi $100B companies in data category is inevitable. and that happening over the next few years is more likely than happening in the next decade.

I always thought it was just me for whom Codex was MUCH faster than Claude Code. But today, the head of AI infra at a $10b company confirmed that Codex is indeed much faster “due to tool call speed” (his take). Claude Code team, do you know / see that CC is much slower than Codex? It “thinks” for much longer (maybe loading tools / skills?) I’m sad to say that it has really hurt my usage of CC, and I almost dread using it because it just takes too much time compared to the same query in Codex.

At 11k employees, our AI costs are going up. Which model & harness should we use to lower cost but also retain great quality? We didn't want to blindly trust public benchmarks. So we ran a comprehensive evaluation on our tasks, code base, infra. It's been produced by more than 3,000 software engineers, spans 3 hyperscalar clouds and many languages and tasks. The results are surprising. We find that for the SAME mdoel, the choice of harness can significantly save costs (~2x). We also find that GLM 5.2 performs extremely well. We run Omnigent in front of these and can easily multiplex different harnesses and models for different tasks. Check it out: databricks.com/blog/benchmark…

Provisioned Throughput is a new serverless product from @togethercompute w/ guaranteed TPMs. We created this for mission critical enterprise class applications. It functions like dedicated capacity but with similar per token pricing as serverless. Available for @MiniMax_AI and @Zai_org models immediately (and more to come soon).

@ashl3ysm1th @vermilionfund @DominicMadori @TechCrunch Congrats, Ashlee!

Today @vermilionfund closed our $25M second fund. Still backing technical founders building for technical users in AI infra, security, and dev tools. 95% of my Fund I investors came back for Fund II. That's the number I'm most proud of. @DominicMadori from @TechCrunch wrote about our story techcrunch.com/2026/07/08/sol…

Factory (@FactoryAI) CEO Matan Grinberg (@matanSF) says "In the next 12 months, 90% of tokens will be going to open models." Enterprises are already shifting rapidly.. "At the beginning of the year, <1% of their tokens were going to open models." "Around March, it crossed 1%... & in May it crossed 10%." "These open models are incredibly performant, incredibly cheap, incredibly fast." While frontier closed models aren't going away: "They'll still have a place, but I think that place will be shrinking—at least in token share." And by the end of 2026? "At least in the enterprise, we'll probably cross the 50% threshold."

@xdotli Well said, Xiangyi!

you can build anything you want you can't build everything you want focus is #1 thing at a startup

@ZacharyDeWitt Compute/AI Inference and Data Training

AI (OpenAI, Anthropic), defense (SpaceX, Anduril), and fintech (Stripe, Ramp) have proven they can absorb billions in VC funding and still return 5x+. The debate now: what's the next category that can support this much capital?