Model APIs ate the coding harness — own policies, memories and evals instead
rseroter · x · 2026-11-20
Model APIs have absorbed most coding-harness behavior — thinking state, implicit caching, structured outputs, server-side state, hosted loops and sandboxes — while open specs (AGENTS.md, SKILL.md, MCP) standardized the outer contract across Claude Code, Codex and Antigravity. Swapping harnesses is now near-zero cost, so teams debating frameworks miss the point. What you should still own: policies (sandboxes, approvals, caps), code (repos, worktrees), memories (rules, skills, repo context) and evaluation (stop hooks, log audits). Understanding and influencing model behavior is where engineering value lives.
More from coding & agent
- DeepORG open-sources an AI agent that runs across your org's databases, APIs and code — dosco · 2026-11-22
- If an agent can still move money after its prompt says no, the boundary is in the wrong place — ExecLayer_io · 2026-11-22
- AutoRAGBench: LLM agents build RAG systems beating the best published human designs — omarsar0 · 2026-11-22
- Stop @-ing your agent in Slack — let Jev decide when it should respond — DavidWells · 2026-11-22
- Intent adds notch launcher, Claude Code and Codex session import — Wattenberger · 2026-11-22
- rakyll explains AX: a low-opinion job orchestrator that agents control themselves — rakyll · 2026-11-22