Tomer Tunguz · 实时热榜

The harness now moves the coding benchmark more than the model does. Endor Labs' Agent Security League found GPT-5.5 scored 61.5% functional correctness in Codex & 87.2% in Cursor, & Claude Opus 4.7 scored 87.2% in Claude Code & 91.1% in Cursor. Input tokens are 86-98% of OpenRouter volume, so the harness controls most of the bill through cache discipline. First-party co-design buys real cache hit rates, but a third-party harness can match them.

A year-old GPT-OSS-120b still serves 36% of Claude Opus 4.8's daily token volume on OpenRouter because the inference market has segmented across cost, speed, & accuracy. GLM 5.2 now out-serves the frontier at 495B tokens a day, & Opus 5 shipped to contest the open-weight field. Mixture-of-experts lets a 118B model decode at the cost of a 26B model, moving the boundary of the local segment.

Google Cloud reported 82% YoY growth to $24.8b in Q2 2026, began recognizing revenue from TPU system sales delivered to customer data centers, expanded operating margin to 35.6%, & grew backlog to $514b, with its growth rate converging toward NVIDIA's Data Center segment.

AI engineering productivity outcomes cluster in three tiers: a mean of 20-46% from distributing an IDE, a frontier of 2.5-3x from companies building the operating layer around agents (Replit, NVIDIA, Amplitude, Anthropic), & a factory tier at 8x+ where agents operate as first-class organizational units (Nubank with Devin). The gap is not the model but the operating discipline built around it.

Open-weight models have repeatedly reached equivalency with closed frontier models, first with DeepSeek R1, then GLM-4.6, GLM-5.2, & Kimi K3. But open source has not materially led; closed models such as GPT-5.2-class systems & Opus continue to create step-change moments.

Slate Auto's $24,950 customizable pickup truck & Thinking Machines' Apache-2.0 Inkling model pursue the same strategy: ship a deliberately unremarkable, general-purpose base cheap & let customers build value through customization. That business model funds open source in AI.

The harness, the software wrapping the model, is becoming the strategic asset. It decides what data flows in, what gets logged, & what gets used to train the next model.

AI customer retention sits between mobile games and social networks. Frontier models keep the crown for about 41 days, and the price of a given level of intelligence falls roughly 10x per year, giving buyers more leverage every cycle.

How AI reshaped the software stack, venture stage definitions, and the firms that win over three years.

Ollama makes open models effortless to run locally on a laptop or in the cloud with the same experience. 8.9m developers, 67,000 integrations, & partnerships with every major model lab & hardware vendor. Theory led the Series B.

A working memory architecture for AI agents : preflight retrieves the right skill from a long-term library, a local model executes, & a watchdog reads the trail overnight to update the library.