Intel researcher: AI agents now run all my ML training experiments, but over-engineering persists
JFPuget · x · 2026-10-19
Intel researcher JF Puget shares his evolving reliance on AI agents: they now handle all of his training experiments, and while he still reads agent harness code, he suspects that won't last long. He currently uses Codex with astra max, having been burned by Claude Code's TUI (not the models or output quality).
His main complaint, tested across several models and harnesses:
- Over-engineering: coding agents make changes he never asked for, forcing constant rollbacks
- Dead code: agents keep code "for compatibility" when compatibility was never a concern
- The result is harder-to-read code and wasted token consumption
His verdict: software engineering is not solved — it's largely automated, but humans still need to drive. As evidence, he points out that the Claude Code team only recently discovered that zipping text enables faster transmission; if SWE were truly solved, that would have been done months ago.
More from coding & agent
- State machines for agent logic are extremely slept-on: LEGO bricks agents can rearrange — DavidKPiano · 2026-10-19
- Data preparation is architecture: a six-step pipeline shared by fine-tuning and RAG — Puzzleheaded_Box2842 · 2026-10-19
- Inspired by Booking.com's MCP playbook, author launches a Query Agent MCP server — CShorten30 · 2026-10-19
- Booking.com's platform team distributes MCP servers so internal teams can build tool-calling agents — CShorten30 · 2026-10-19
- Hunting for an agent that auto-tops up LLM API balances across providers — huangyun_122 · 2026-10-19
- LangChain CTO to join fireside chat on turning production agent traces into better models — hwchase17 · 2026-10-19