Models improve faster than tinkerers: vanilla Codex users get the state of the art
pvncher · x · 2026-10-06
- thdxr points out a weird inversion with LLMs: models improve faster than the people tinkering with them.
- Custom workflows and setups are increasingly solving problems that no longer exist; the person naively using vanilla Codex is more likely experiencing the state of the art.
More from coding & agent
- Kabir returns to Stripe, doubling down on agentic payments infrastructure — jeff_weinstein · 2026-10-06
- Trained agentic context management: 8K-context small model matches GPT-5.4 at 1M on OOLONG — xennygrimmato_ · 2026-10-06
- Granola now auto-captures screen shares in meetings, not just what's said — soleio · 2026-10-06
- One Agent's Silent Retries Were Eating 60% of His API Spend — Artistic-Earth8997 · 2026-10-06
- Agents Write PRs Faster Than Humans Review: One Team's Fix — nateritter · 2026-10-06
- Spawn turns Discord feature requests into shipped features in under 12 hours — majidmanzarpour · 2026-10-06