Addy Osmani on coding agents: separate adversarial-review models add little over same-session prompts
addyosmani · x · 2026-10-08
David Cramer (@zeeg) argues that adversarial review in AI coding workflows doesn't require a separate model or even a separate context window: just ask within your existing session for similar results. Drop the "compaction paranoia" and don't count on "fresh context" helping — it often won't.
Addy Osmani agreed, saying he's had similar feelings about the net value of using a separate model for adversarial review, and would like to see more research on the topic.
Related event: Devs Use Rival AI Models to Review Each Other's Code(2 posts)→
More from coding & agent
- ChatGPT PM: AI-era product management means steering probabilistic systems — AarushSelvan · 2026-10-08
- Anthropic engineer's favorite Claude PM trick: auto-find top 10 users and schedule feedback calls — _catwu · 2026-10-08
- GitHub Copilot embraces local AI at Microsoft Windows launch, touting llama.cpp — film_girl · 2026-10-08
- Hyper-Extract: Open-Source CLI Turns Documents into Hypergraph Knowledge with LLMs — tom_doerr · 2026-10-08
- OSHBuilt's bet: MCP servers become the standard interface that lets AI agents source US manufacturing — hudzah · 2026-10-08
- Spacebar launches multiplayer AI workspace where a whole team directs one AI coworker — Scobleizer · 2026-10-08