Discussion on Codex Auto-Compaction and Settings
soumitrashukla9 · x · 2026-07-12
This repost focuses on the user experience and setup details of Codex, highlighting two main points:
- Context Window Counter: Some users don't know how to enable this counter, and the author suspects an update might have disabled it by default. As Codex settings become increasingly complex, the author hopes for natural language settings search similar to Android, allowing users to simply type "how do i turn on context window counter?".
- Auto-Compaction Threshold Change: The quote mentions that Codex's baseline auto-compaction has increased from a historical 262K to 353K. For long-running PC/browser automation tasks, this change is crucial because such tasks generate massive amounts of accessibility tree tokens; over-compacting leads to more severe information loss and slower performance.
The quote also adds that although GPT 5.6 features a 1M context window, it "shouldn't be used" on Luna as performance degrades.
Related event: OpenAI Codex Raises Auto-Compaction Context Baseline(3 posts)→
More from coding & agent
- huggingface_hub 2.3.0 cuts 1M-file listings to 20s, adds live Bucket following — lhoestq · 2026-10-27
- Nous Research ships Advanced Delegation plugin letting Hermes Agent route subagents to different models — NousResearch · 2026-10-27
- A dentist with a 3D printer discovered OpenAI Codex — and never looked back — gabrielchua · 2026-10-27
- AutoResearch agent produces weird 'forking' loss curves, and the team dug into the science — ZimingLiu11 · 2026-10-27
- Stanford's PaperFormalizer turns research papers into Lean-verified versions with author in the loop — StanfordUniversity · 2026-10-27
- Hugging Face TRL makes score centering + masked importance sampling default in async GRPO — _lewtun · 2026-10-27