Six config changes to give GPT-6 Astra more agency in Codex
daniel_mac8 · x · 2026-09-07
Dan McAteer shares field notes from two days of heavy GPT-6 Astra use in Codex, plus six config changes to maximize its agency.
- Enable persistent context: Astra's experimental context-management feature is off by default. Adding [features] contextmanagement.experimentalmode = true to config.toml lets it save notes across context windows and search earlier messages and tool calls, avoiding the fragile single-summary compaction that loses information on long tasks.
- Paid follow-ups cover multi-agent setups, an explore-then-implement workflow, and an "Astra Field Kit" with configs, prompts, a configuration builder, and a checker.
His core thesis: Astra's limits now feel like the user's own imagination and goal clarity, and Codex setups carried over from the GPT-5.x era should be reworked around the new features.
Related event: Six Ways to Get the Most Out of GPT-6 Astra in Codex(2 posts)→
More from coding & agent
- Shai-Hulud npm payload reemerges after 111 days, slipping past npm's malware scanning — jedisct1 · 2026-09-07
- This Week's AI Must-Reads: OpenAI's Research Acceleration Report and Broadcom's $16.7B AI Chip Quarter — VibeMarketer_ · 2026-09-07
- Scanning 5,600 Lovable/Bolt apps found 400 leaked secrets and 175 leaking customer data — Warm-Reaction-456 · 2026-09-07
- On-device Android agent with Gemma 4 E2B hits 2.6 tok/s live vs 11 tok/s on replay — HowDevelop · 2026-09-07
- Dev open-sources a dependency-free LLM eval harness that catches regressions, not averages — zahidsaim23 · 2026-09-07
- Coding agents on 8GB VRAM: little-coder vs barebones just Pi for small models — Botoni · 2026-09-07