Enterprises should own business knowledge, not models, as frontier agents double every 6–7 months
hamostaf04 · x · 2026-07-26
The article argues that enterprises are owning the wrong thing if they focus on owning a model instead of owning business knowledge.
- Frontier model capability is compounding quickly: on METR’s software-task benchmark, the length of tasks a frontier agent can complete with 50% reliability has roughly doubled every 6–7 months.
- Costs are also dropping fast: Stanford’s AI Index found GPT-3.5-level querying got more than 280× cheaper in about 18 months.
- The author says the real moat is company-specific knowledge—how decisions are made, which edge cases matter, and what outcomes count as success.
- The recommended loop: record work and outcomes, capture corrections and edge cases, convert them into reusable examples/rules/checks, then carry that knowledge forward to each new model.
- A claims-processing example shows how an expert override becomes a durable rule and quality check for future models.
More from Companies & People
- Anthropic employee says 6 months there have been “incredible” and shares a hiring link — EricBuess · 2026-07-26
- OpenAI signs open-weights letter as model serving at scale gets harder — steipete · 2026-07-26
- Sonnet 5 spent weeks autonomously fixing builds, filing issues and reading README files — Sauers_ · 2026-07-26
- Teams can run evals, but still can’t tell if a release is really better — hwchase17 · 2026-07-26
- Pupper Build Day mixes new builders with Unitree humanoids and students — DynamicWebPaige · 2026-07-26
- Local AI ecosystems are shifting assistants from chatbots to action layers — TheTuringPost · 2026-07-26