Henry Yin

OpenClaw was the most important application signal of 26Q1: it did not invent new Agent technology, but combined local permissions, messaging apps, long-term memory, scheduled tasks, and tool loops into an “iPhone moment for AI Agents.” Henry says its GitHub Stars surpassed React’s 10-year accumulation in roughly 60 days. The core shift is that “AI comes to your life, rather than you going to find AI”; it validates mass demand for personal Agents while pointing Claude Code, NVIDIA, and OpenAI toward the next product roadmap.

This quarter, Anthropic went from a “technically respected challenger” to a platform-level rival capable of directly threatening OpenAI, driven less by benchmark leadership than by the Claude Code product–revenue flywheel. The episode cites multiple figures: Henry says Anthropic’s revenue is growing at roughly $5B per month; 曼琪 says Claude Code ARR rose from $900M in December 2025 to $1.9B in early March 2026, while Henry separately puts it at roughly $2.5B ARR. OpenAI rose from $21.4B to $25B; roughly 70%-75% of Anthropic’s revenue comes from B2B and API.

The Opus 4.6 versus GPT-5.4 matchup shows what really decides the coding Agent race: not how much good code a model generates in one shot, but who can more reliably complete the full engineering loop of planning, debugging, testing, reading logs, and opening PRs. Opus 4.6 has one million context and reportedly can work continuously for roughly 15 hours; GPT-5.4 scored 75% on OSWorld, above the 72.4% human score cited on the show, and is often considered stronger at pure coding. Developers, however, are using Claude Code to plan and Codex to execute, creating the awkward split in which “Claude Code is the master, Codex the slave.”

High-frequency, multi-turn Agent reasoning has made price a decisive variable in model share, giving Chinese models a real distribution window inside the OpenClaw ecosystem. MiniMax M2.5/M2.7 cost roughly $0.2/$1.2 per million input/output tokens, versus $5/$25 for Opus 4.6—around a 20x gap. The show estimates monthly costs could fall from $200 to roughly $15 after switching. “A 20x cost gap multiplied by call volume” is enough to put Step, MiniMax, Kimi, GLM, and Xiaomi among OpenClaw’s top users on OpenRouter.

AI self-evolution has moved from a science-fiction premise to a small-scale engineering loop that can actually run, though it still depends on humans defining the objective and search space. Andrej Karpathy’s AutoResearch runs for roughly 15 minutes per iteration and nearly 100 experiments per night, finding more than 20 effective improvements and cutting training time for a GPT-2-level model by roughly 17%-20%. It is best suited to tasks with clear metrics, fast feedback, and automatically verifiable outcomes—performance, kernels, database queries—not autonomous selection of research direction.

The infrastructure investment cycle is shifting from training to inference, and Agents will pull along GPUs, CPUs, storage, and security infrastructure at the same time. Henry relays that the Vera Rubin platform could deliver roughly 3-5x higher inference performance and a 10x reduction in token costs, while Google’s KV Cache quantization work could reduce storage requirements to one-sixth of current levels. Because “everything is becoming a computer,” an Agent that writes code must also launch a sandbox and execute it, potentially creating CPU shortages; one investment thesis he heard was to seek relatively pure exposure such as Arm.

Companies have begun converting AI gains into layoffs and a reordering of talent, raising the bargaining power of top performers while potentially compressing middle-tier jobs and traditional SaaS margins. Henry cites Amazon cutting 16,000 jobs, Block cutting 40%, and Meta planning to cut roughly 20%, or 15,000 people, while lifting AI capex to roughly $65B. 曼琪 observes that Chinese startups are also moving from “one top-tier person plus second- and third-tier staff” to “superstar talent plus Agents.” This is not simply cost-cutting: capital, revenue, and Bay Area housing could all become more polarized, even prompting discussion of a “token tax.”

🔗 Original source & video: 156: Henry’s “AI Quarterly 26Q1”: OpenClaw, the Three-Way OpenAI–Anthropic Showdown, and Self-Evolution

The significance of Gemini 3 Pro is not that it leads on any single benchmark, but that Google has finally combined long context, cost efficiency, multimodality, generation quality, and developer access into a product moment that reaches SOTA across different dimensions. The guest attributes the comeback to long-term co-design: vertical integration across TPU, model infrastructure, and applications, with Workspace and multiple hardware surfaces continuously feeding data back into the system; Google Labs then used editors, writers, and people skilled at creating viral posts and works to refine the product’s distribution. More importantly, the team says pre-training still has “no war in sight”—this is not a one-off catch-up.

GPT-5.2’s benchmark surge only makes sense on the cost curve: GDPval rose from 38.8% for GPT-5.1 to 77.9%, while ARC-AGI-2 climbed from 17.6% to 52.9%, even as the actual cost of equivalent tasks fell. The guest sees this as higher “intelligence per token,” not merely brute-force test-time scaling; but GDPval’s task mix and document entry points may still be toy-like, and the model can only answer how many r’s are in Garlic probabilistically—“AGI still has a long way to go.”

Foundation models will continue to internalize general-purpose Agent scaffolding, but will struggle to absorb the heavy tail created by proprietary enterprise workflows. Model vendors can train models on mock tool interactions to develop “their own shortcuts,” but enterprise-specific, non-public data will leave a persistent gap. Agents such as Cursor, with access to real scenarios and users, can in turn push models to improve coding, debugging, and multi-step capabilities. “Models and Agents make each other better”; the opportunity worth backing is the net-new market that continues expanding even after models absorb legacy functionality.

Precur is betting on a trainable tool layer: tools are no longer static APIs, but systems that remember “where they themselves have failed,” turn failure trajectories into enterprise assets, and improve autonomously. Its code-driven tool aims to reduce context pollution in long tasks; in early customers’ blind evaluations, it delivered roughly a 12% improvement over incumbent tools while reducing latency and expanding coverage, without requiring customers to replace their existing Agents. The end market is neither To C nor To B, but “To Agent”—a market served by the millions of Agents created by people through vibe coding.

The RL startup opportunity is shifting from “squeezing more out of data” to competing for interaction, reward, and high-fidelity environments, across three layers: RL environments, RL as a Service, and high-value vertical applications. The environment layer will become the “practice ground and exam center” for future Agents, but replicating Jira is no longer a moat; cybersecurity offense and defense, robotics, and autonomous-driving simulation require domain expertise. The service layer serves enterprises that cannot build RL in-house, while the applications layer targets drugs, trading, and science discovery. Moe Capital has invested in Preference Model, while Periodic Labs represents the third category.

An Agent’s model choice is first a question of cloud, discounts, and ecosystem, and only then a question of leaderboards: migrating enterprise data to another cloud often takes 1–2 years, and existing Azure and GCP relationships can outweigh short-term model leadership. Claude’s advantage is not just coding, but the closed loop formed by MCP, programmatic tool calling, Claude Code, and its developer paradigm. Real-time Agents may instead prefer low-latency models such as Gemini Flash and OpenAI mini. Qwen and other Chinese open models offer size choice and room for ablation, and teams will also post-train on open models; policy restrictions and value concerns among US enterprises still leave room for an “American DeepSeek.”

The next Agent capability race will center on native long-horizon reasoning, multimodality, and open models that can be post-trained—not merely another fragile layer of scaffolding. Developers need open foundations that have seen abundant Agent traces during training, can directly understand dirty PDFs, images, and mixed modalities, and can plan, act, and correct themselves across tasks spanning dozens of steps. Even if models internalize some context engineering, framework usage data will continue feeding back into model training; the guest’s view is that the serviceable market will expand faster than the portion absorbed by foundation models.

🔗 Original source & video: 146: Behind Gemini 3’s Comeback, What Models Agents Need, and the RL Startup Opportunity | A Conversation on Bay Area Trends with a Former Google Entrepreneur and a Silicon Valley Investor

OpenAI is compressing Agent building, deployment, evaluation and optimization into a standardized production line. Agent Builder orchestrates classifiers and if/else logic through drag-and-drop workflows; ChatKit handles frontend delivery; datasets, trace grading, prompt optimization and reinforcement fine-tuning form the feedback loop. Henry’s summary: OpenAI “trained itself in every martial art, and now wants to teach them to developers everywhere.” This lowers the barrier to enterprise adoption, but may also standardize away part of the performance moat that once belonged to Agent startups.

The core asset of Apps in ChatGPT is not a new UI, but operating-system-level distribution to 800M weekly active users. Unlike most GPTs in 2023, which were little more than “wrappers around prompts,” the new Apps SDK is built on MCP and adds OAuth-style authorization, external tools and interactive components, allowing apps such as Canva and Figma to deliver a fuller product experience. Revenue sharing remains experimental; the idea of a future 30% take rate is only an iOS-style analogy. Developers also face an asymmetry in data and retention: “It’s a bit like building a house on someone else’s foundation.”

Agent Builder exposes the real tension between OpenAI’s commercialization strategy and its AGI path. The AGI vision is for models to absorb hard-coded human workflows, while Claude Code-style general Agents improve directly as the underlying model gets better; Agent Builder instead turns workflows into diagrams. The trade-off is safety, explainability and something large enterprises can buy today. Henry called it “a very pragmatic pivot”: OpenAI wants autonomous Agents, but also needs current enterprise revenue to support a valuation of about $500B and even larger ambitions.

Agent Tooling could jump from a roughly $20B–$30B DevTools submarket to a long-term $200B–$500B market. The guest starts with a global software market of about $650B and DevTools at a low-to-mid single-digit share, then cites the view that software could expand to $10T as it absorbs services. If the tooling layer captures about 5% of the new market, it could grow 10–15x. The biggest opportunities are not tool catalogs, but products that control identity, communications, observability or evaluation and build data loops where “the more you use them, the higher the pass rate and the lower the cost.”

The tooling ecosystem is moving from “can call tools” to “can complete tasks,” with Composio as a representative case. It uses Agents to write and repair MCP servers, then positions Rube as “one MCP server to rule them all,” allowing Cursor to select tools from hundreds of servers based on the task. This eases context and token pressure while turning reliable execution into a product. MCP is better suited to high-success-rate critical workflows, especially those involving writes; browser use fills the long tail of sites without open APIs. The two are more likely to coexist than converge.

Voice has already produced a measurable infrastructure multiplier: LiveKit went from carrying about 1M voice calls per day to 20M within a year. It does not primarily make voice models; it provides real-time audio/video, turn detection and orchestration. Its customers include Character.AI and Grok, and it says it supports roughly 25% of 911 traffic and helps save one life on average each week. Speech-to-speech is seen as the end state, but the STT→LLM→TTS cascade still has a clear commercial runway because it offers more control over guardrails, cost and behavior.

Evals and memory are not ancillary features; they are the control layer that moves Agents from demos into production. Even a large customer may release a voice Agent after engineers test only 3 or 4 calls because labeled data is expensive, teams cannot agree on evaluation sets, and subjective, complex tasks are harder to verify than coding or math. OpenAI’s $1.1B acquisition of Statsig shows that A/B testing, phased releases and metric loops are being built in. Letta’s “sleep-time compute” represents another path: when no one is interacting with the Agent, it spends tokens organizing long-term memory and continuously converting state into reusable capability.

🔗 Original source & video: 137: Agents Are the Opportunity—and So Are the Tools That Build Them | From OpenAI Dev Day｜Agent #6