Anthropic engineer: build systems that prompt themselves — 90% of our engineers already do
Paimaamu · x · 2026-10-27
A widely circulated thread centers on an Anthropic engineer's claim: "You're not supposed to prompt Opus 5.5 — you're supposed to build a system that prompts itself. At Anthropic, 90% of our engineers have already built such systems." She demonstrates in 30 minutes how Anthropic builds agents with memory, self-error-correction, and per-run improvement; the quoter calls it better than any $1000-$2000 agent course.
The reposter adds that all the harnesses and brand systems he built earlier work well with Opus 5.5 and other models: "stop chasing models, chase clear outcomes" — system designers are the real winners of the AI era.
More from coding & agent
- A daily two-checkpoint routine for managing AI agents at work — RachelVT42 · 2026-10-27
- "Is it time to fork Zed?" Community pushback over the editor's AI direction — gnukeith · 2026-10-27
- Reddit hands-on: Qwen3.5 9B struggles at agentic work while 35B A3B handles it — Nautisop · 2026-10-27
- JetBrains gifts researcher a flickering agent tracker, cooking his work-life balance — mervenoyann · 2026-10-27
- Developer uses Claude Code and Codex to find a possible planet — CodeByPoonam · 2026-10-27
- Fake a '200 OK' and the agent just keeps going: an agent reliability experiment — ngxson · 2026-10-27