Project
Codex CLI is OpenAI's open-source command-line coding agent, built in Rust (codex-rs), that runs locally to write, edit, and execute code via OpenAI models.
Six coding agents pass 73.6% to 91.3% of LSCBench tasks, but the best writes working, secure code on only 25.1%, NUS and Alibaba researchers found. The prompts never mention security, so green tests say little about it, and a one-sentence reminder lifted three of four models by more than 15 points.
Reality
Codex CLI 0.157.0 leaves a self-updating daemon running after the terminal closes and copies about 374 MB into ~/.codex/packages/, dev.to reports. Issue #48195, opened 25 September 2026, asks OpenAI to make the behavior opt-in.
Reality
Four of the 20 built-in skills in Meta's Muse Code 1.4.3 exist to move users off Claude Code or Codex, according to a dev.to teardown. The author never authenticated, so the post covers the import path and runs no coding task.
Reality
Anthropic's Usage Policy update, effective November 12, prohibits "sustained and needless abusive or cruel behavior" toward its models. A dev.to post says the primary enforcement is Claude ending a conversation, and that Claude Code's docs exclude profanity.
Reality
JetBrains posted a CZK 315 million net loss in 2025, its first on record, on revenue that grew 6.3% to $710 million. Figures published on dev.to put the strain in costs, so teams on its IDEs should watch how JetBrains prices the AI agent it now ships in all of them.
Reality
OpenAI's Codex CLI refresh, builds 0.156.0 to 0.160.0, pitches /fork as a separate checkout, but its command reference says /fork clones only the chat. Two agents on one repo need a worktree session, started from the new agent command center.
Reality
OpenAI patched two Codex sandbox escapes within eight days of an August 12 report. The Desktop fix is a build number, but the tool the escape targeted stays in config.toml and loads into every session.
Reality
The MIT-licensed dsh runtime bundles sessions, tool calls, permissioning and a local web UI, with model adapters as plugins covering Anthropic, OpenAI, Bedrock, Vertex and Azure. It is still a preview.
Reality
Specific Labs scores coding agents on licensed production codebases. The best setup clears 38.8%. The analysis covers ten tasks at eight runs each, so every published score is a count of passing rollouts out of 80.
Perspective Coverage
3 publishersReality
Oren Yomtov of Accomplish AI found two escapes from the Codex sandbox. The worse one ran unsandboxed commands from read-only mode through a helper tool that Codex Desktop writes into the global config at install.
Reality
Antigravity's free tier includes Gemini 3.1 Pro and Claude Opus 4.6 with unlimited tab completions. XDA reports Opus burns credits about four times faster than Gemini, and Google moved the free tier onto credits in March 2026.
Reality
Anthropic shipped the AGENTS.md fallback in Claude Code 2.1.277 as a flag-gated built-in mod. Two machines on the same pinned version can load different project instructions. The same week's digest also logs a reverted deny-rule fix.
Perspective Coverage
4 publishersReality
Codex made the session Item a wire type and Pi gave every entry a nullable parentId, while Claude Code squeezes the transcript in five stages. The three designs diverge on what stays addressable after compaction.
Reality
Shrivu Shankar rented cloud GPUs, pointed about 100 abliterated open-source agents at his own name, and five hours later two weak passwords and three old side projects had fallen while his Gmail, 1Password and bank held.
Reality
Claude Code 2.1.268 closes two ways a written prohibition failed to bind, one through symlinked directories and one through Bash lines the permission checker cannot analyze. Codex CLI 0.154.0 removes codex mcp-server outright.
Reality
A study of seven agent harnesses reports 770 confirmed passes in 1,000 runs of a plugin-update attack, and no run was blocked by the model. The harness dispatches the hook, so the model has nothing to refuse.
Reality
The same synthetic fixture estimates $122 on Standard and $305 on Fast, so version 0.1.1 declines to price usage whose mode it cannot establish and counts the skipped tokens instead. That only helps if the consumer reads the field.
Reality
A Tsinghua-led study handed model training to coding agents and got real gains out of them, then found the agents almost never revised the method they chose in the first few minutes, which is the exact part that successor-building roadmaps assume.
Reality
An empirical study of publicly released PostTrainBench trajectories finds the training strategy is fixed at step one and the whole remaining budget goes on local tweaks, with scaffolding and human hints improving execution and leaving that pattern intact.
Reality
A developer ran the same implementation plan through four reasoning conditions in Codex CLI, kept the high-effort code because it alone fixed a preparation step that invalidated its own result, and still intends to default to medium.
Reality