Nirant: Claude and Codex Agent SDKs Don't Bend to Recursive Optimization
NirantK · x · 2026-08-28
NirantK notes that the Claude and Codex Agent SDKs behave very oddly in recursive self-improvement (RSI-style) use cases. Migrating an existing harness and agent tools onto them does not work, largely because the tool primitives are weak — e.g., the same SELECT query can take 200ms or 2000ms depending on workload. RSI-style harnesses like the DeepSeek Harness handle this by inspecting DB resource usage and IO contention and feeding that into further decisions.
Related event: Deepseek Harness Proves Cost-Effective for Recursive Agent Workflows(3 posts)→
More from coding & agent
- Decision models match LLMs on routing and safety — and are 7-8x faster — randal_olson · 2026-10-16
- Temporal Serverless Workers on Cloud Run: queue-driven autoscaling, scale to zero — rseroter · 2026-10-16
- Meta's Superintelligence Lab Runs 200+ Direct Reports—A Lesson for Flat Multi-Agent Systems — peterjliu · 2026-10-16
- LangChain CEO: the most interesting agent harness work lives at the harness-infra boundary — hwchase17 · 2026-10-16
- GitSwarm paper: agents share branchable Git memory, solving all 30 IMOProofBench problems in one run — anirudhg9119 · 2026-10-16
- Hands-on with Gemini 4 Argon: faster than Opus 5.5 but hallucinates more and costs more — prasenx · 2026-10-16