Mid-run prompts add reproducibility and leakage-risk columns to Apodex research table
ChrisGPT · x · 2026-09-27
A follow-up to the Apodex research demo: while the agent was running, ChrisGPT added instructions to keep only results with public data/code or enough methodological detail to reproduce.
He also asked for leakage risks and tests to disprove each claim — those columns appeared in the finished comparison table, showing how mid-run constraints shape agent output.
More from coding & agent
- Dev builds a zero-LLM Windows harness so Jev computer use can do multi-step tasks — airesearch12 · 2026-09-27
- One prompt, 6 hours of Claude Code: a 6-minute JCS-style interrogation film — nt_coco · 2026-09-27
- Vibecoding at Five: Critics Declare It Failed While Adopt 'Tell the AI to Fix My Linux' Skyrockets — teortaxesTex · 2026-09-27
- How expensive are silent agent failures, really? Reddit weighs debugging costs — Sensitive-Parsnip-12 · 2026-09-27
- LLMs Are 'Stupid' Yet Productive at Code: Brute Force Beats Intelligence — ivan_bezdomny · 2026-09-27
- Devs on Reddit debate: what do you actually do after your agent eval catches a failure? — Sensitive-Parsnip-12 · 2026-09-27