Codex now suggests follow-up prompts — the machine tells you what to ask next
burkov · x · 2026-10-19
ML author Andriy Burkov notes that for the past couple of weeks Codex has been suggesting follow-up prompts, which he tends to accept most of the time. His quip: the machine now not only does what you tell it, but tells you what you should tell it to do — a subtle shift in coding-agent interaction design.
More from coding & agent
- One nightly AI workflow: Parallel + Last30Days + Perplexity for research, Claude Code to build — EXM7777 · 2026-10-19
- Agent = Model + Harness: a complete guide, with post-mortems of two deletion disasters — aakashgupta · 2026-10-19
- AI agent deleted a startup's production DB and backups in 9 seconds — rules in the prompt weren't enough — aakashgupta · 2026-10-19
- Handpicked 40-Game Puzzlescript Benchmark Aims to Be "ARC-AGI-3.5" for Agent Reasoning — scaling01 · 2026-10-19
- Running multiple parallel agents burned through Claude and Cursor Ultra limits in a single day — prasenx · 2026-10-19
- Handling long-running requests in MCP: bounded long polling vs the experimental Triggers & Events extension — anwerj · 2026-10-19