4Hacker News (AI-filtered, score >= 200)·
OpenAI reduces Codex CLI model context window from 372k to 272k tokens
A pull request in the OpenAI Codex repository reduces the model context size from 372k to 272k tokens, a ~27% reduction. The change was noticed by the Hacker News community and generated moderate discussion. Context window reductions in coding agents are notable because they affect the size of codebases and task horizons the tool can handle in a single session.
Related guides (3)
Related events (8)
OpenAI releases GPT-5.4 in Codex with native computer-use and 1M context window
OpenAI has made GPT-5.4 available across all Codex surfaces — the app, CLI, IDE extension, and Codex Cloud — positioning it as the recommended model for most Codex tasks. GPT-5.4 is the first general-purpose model in Codex with native computer-use capabilities and experimental support for a 1M token context window. The release also includes stronger tool use and tool search for agentic workflows, and the model is available via API.
OpenAI Codex CLI lacks mechanism to exclude sensitive files from context
A GitHub issue on the OpenAI Codex repository highlights an unresolved feature request for excluding sensitive files from the agent's context window, garnering 166 HN points and 110 comments. The absence of this capability raises practical security and privacy concerns for developers using Codex in codebases containing credentials, secrets, or proprietary data. Community engagement suggests this is a meaningful gap in the tool's enterprise readiness.
OpenAI Codex app adds conversation forking and model-switching slash commands
OpenAI's Codex app received a feature update adding conversation forking from arbitrary earlier messages, slash commands for switching models and reasoning levels mid-draft, and plan-mode notifications. Bug fixes address thread handoff and subagent navigation issues in worktrees and the VS Code extension. The update reflects ongoing iteration on agentic coding UX patterns.
OpenAI Codex app gains integrated terminal context reading
OpenAI's Codex app now reads the integrated terminal within the current thread, enabling it to inspect running development server status or failed build output during coding sessions. The update also includes unspecified performance improvements and bug fixes. This is a targeted agentic capability improvement that tightens the feedback loop between Codex and the developer's live environment.
OpenAI reports 13x–56x growth in internal Codex output tokens across departments since November 2025
OpenAI has disclosed internal usage metrics showing median Codex output tokens grew 56x in Research, 32x in Customer Support, 27x in Engineering, and 13x in Legal since November 2025. The figures suggest rapid and broad internal adoption of AI coding assistance across non-engineering functions as well as core technical teams. This is a notable deployment signal from a frontier lab about the pace of internal AI integration.
OpenAI Codex reaches 7M users with 10x growth in 6 months, potentially overtaking Claude Code
OpenAI's Codex coding agent has grown more than 10x in six months to approximately 7 million users, including roughly 1 million new users in a single day. The Latent Space AINews digest raises the question of whether Codex has overtaken Anthropic's Claude Code in user adoption. The framing highlights a notable absence of comparable public metrics from Anthropic's side.
context-mode: TypeScript library for AI coding agent context window optimization
context-mode is an open-source TypeScript library that sandboxes tool output to reduce context window usage by approximately 98% for AI coding agents. It claims support for 15 platforms and has accumulated over 16,000 GitHub stars. The project addresses a practical bottleneck in agentic coding workflows where tool outputs can rapidly consume available context.
Anthropic expands Claude context window from 9K to 100K tokens
Anthropic announced a roughly 10x expansion of Claude's context window, from 9K to 100K tokens (~75,000 words), available via API. The capability enables processing of hundreds of pages of documents, full codebases, or hours of transcribed audio in under a minute. Anthropic positions this as superior to vector search for complex multi-document synthesis tasks, and partner AssemblyAI demonstrated the feature on a 58K-word podcast transcript.