Anthropic engineer earning $350K–$950K credits his edge to building agent harnesses
aftahi_ai · x · 2026-11-10
A widely shared thread claims an Anthropic engineer earns $350K–$950K for one skill: building agent harnesses.
- 99% of people hand-build multi-agent systems that keep breaking; he builds loops, sandboxed tools, and systems that turn yesterday's logs into new skills — same model, 10x more reliable agents.
- Five steps: (1) manage loops, context and sandboxing via a harness; (2) separate the brain from tools for 60% faster first token; (3) run agents server-side and recover from crashes using logs; (4) make failures recoverable with retries and replays; (5) turn past execution logs into memories so the system improves over time.
- The tweet says Anthropic calls this process "dreaming" and links to a workshop (promotional in nature, apply with caution).
More from coding & agent
- Kylon goes cross-platform with 150+ agent use cases across 15 industries — testingcatalog · 2026-11-10
- Codex rolls out global dictation using ChatGPT's fast transcription path — gabrielchua · 2026-11-10
- Zero-cost uncensored deep research stack: Ollama + SearXNG + dual-model fact-checking in Hermes — damnGruz · 2026-11-10
- Open-sourcing PAIR: building trustworthy evals on your own data, fair across vendors — kalyan_kpl · 2026-11-10
- Bloomberg's AI coding interview: 30 minutes on a huge Django codebase, still rejected — dotey · 2026-11-10
- Automattic launches Spacefast: hosting and sharing for apps your AI builds — LuoSays · 2026-11-10