I just spent 30 mins with Astra working on 1 issue.
This is a AI post classified by Jev as AI coding agents (a tutorial), kept by the AI Radar because it carries real work, not commentary.
I just spent 30 mins with Astra working on 1 issue. Drained 10% of my weekly usage on the Pro 5x plan. This was the issue: babysitting the PR Codex polls subagents/background tasks once every minute. So it passes the full context as a turn each time it checks on a task. A literal token nuke. Fix - add this to your config.toml: developer_instructions = """ During background tasks or subagent waits, announce the wait once. Then wait silently using the longest supported tool timeout. Let scripts handle polling. Do not wake just to give progress updates or run duplicate checks. Report comple
Posted by Uzi (4.5k followers) 1 h ago · 8 likes · 309 views · view the original post on X. Kept by the AI Radar as AI coding agents.
More AI work like this
- For all the CLI users: Update Codex and turn on Voice mode in /experimental. — @derrickcchoi
- Grok Build v1.0.40 is out with a few practical updates. The main item is miscellaneous… — @blankspeaker
- Indexes your codebase, dependency graph, and git history to give AI agents cited answers… — @tom_doerr
- TIL you hit the "luna reserve" on codex once you run out of credits — @neuralkian
- @maria_rcks — @maria_rcks
- executor v2 was slow so I sent this prompt to Astra and then went out for a walk — @RhysSullivan
- Grok Build v1.0.39 is out with a few practical updates. Standouts include subagents can… — @blankspeaker
- ok now do every local codex / claude chat — @clairevo
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 35.2k posts from 5k X accounts over the last 14 days, 1.5k tools, 19 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-21 00:55 UTC. Full method.