Tomas

PSA for macOS dev/testing: be very careful letting coding agents run tccutil, especially with broad permissions. Codex ran tccutil reset Accessibility instead of resetting permissions for a specific app. That wiped Accessibility grants across all apps, broke input/control permissions, triggered WindowServer/HID weirdness, and made keyboard/clicks appear partially dead until sleep/wake, followed by a full reboot. Thankfully this happened inside a macOS VM and not on my host machine.

Installed Markdown Preview Enhanced to replace VS Code Markdown preview. My org uses PlantUML, so I had to manually download and configure a local JAR in VS Code (already felt questionable). Every time PlantUML updates, some app/window pops up for a split second, steals focus, and disappears. Markdown tooling in 2026 is incredible.

@OUTrum That explains it. I forgot it’s Microsoft.

@tmsdnl because Microsoft? :D

Why is VS Code’s Markdown preview still so bad in the age of Markdown? Slow. Janky. Weird rendering quirks. Not even a copy button on code blocks. Is there actually a better alternative that integrates seamlessly?

Can’t continue a Codex conversation on the go without remoting into my machine. When is the Codex app coming out again?

Running agents that research things, build software, come back with results, and help you make decisions is wild. Is this just what CTOs have been doing this whole time? And they were getting paid how much? The gap between IC and CTO is collapsing.

@fcoury 6/ `[features].codex_hooks` is deprecated. Use `[features].hooks` instead.

Codex 0.129.0 is our and here are 5 New Features you need to know about ✨ 🧵 1/6

AI coding agents have made tinkering absurdly cheap (quality is rough, but improving fast). Forked Rectangle to add shortcut cycling the way I wanted it. Patched Codex to override the shell PATH. Without AI, neither was worth my time. Now it's an evening. And it's still cheating, because I'm an engineer. Wait until my mom can do this.

I built a review script with headless Codex and Claude: 1. Resolve target: worktree, branch, range, or commit 2. Generate the diff 3. Claude reviews it 4. Codex reviews it 5. Claude critiques Codex 6. Codex critiques Claude 7. Adjudicator buckets findings: accepted / rejected / uncertain 8. Accepted findings go to the fixer agent

these are my main use cases for Codex subagents: - spawn 2 subagents to review the PR for regressions or bugs - spawn 1 subagent to explain what the PR does in simple concise bullet points (this is for me to review that the direction was right) - spawn a subagent to verify if the PR review comments are real issues - spawn 1 or 2 suabgents to discuss the best implementation plan and pros/cons post yours 👀

@thsottiaux - better late than never. Codex should let users configure the shell it uses to run commands. Some dev setups depend on the exact shell/PATH startup: direnv/devenv/Nix, Python venvs, Node version managers like fnm/nvm/volta, Git Bash/WSL, and project bin shims. Copying env vars is not enough when login shells reorder PATH or skip project shell setup.

Note to future self and PSA for anyone buying a Mac Mini to sandbox coding agents inside a macOS VM: - 32GB RAM is tight. A macOS VM for Xcode + simulators realistically needs ~16GB allocated to it, leaving little room for the host and additional Linux VMs/containers for agents. - 256GB SSD disappears fast between Xcode, simulators, caches, and VM disks (1TB upgrade already on the way) - Get 10GbE (if your network supports it). Once the Mini becomes a remote VM host, 1GbE screen sharing feels choppy. You cannot upgrade it later. *Does not apply if you are yolo-running agents directly on the host.

MacCeption Setting up a separate macOS VM for Codex agents so they can build/test macOS and iOS apps without touching my main environment.

@adamwathan I create an AGENTS.md and then symlink CLAUDE.md to it. A similar technique with skills too. touch AGENTS.md ln -s AGENTS.md CLAUDE.md

Claude Code not supporting `.agents/` for skills and stuff is annoyingly hostile 🫠

Did you know you can use Computer Use in Codex CLI?

High as a kite, literally. Spotted from Bernal Heights, SF

Nobody talks about how hard it is to ship open source. Prototyping is Fun. Immersive. Exciting. Then comes packaging, docs, CI/CD, bugs you can’t reproduce, edge cases you never imagined. Tedious. Frustrating. Endless. No wonder they say: the first 90% takes 90% of the time, and the last 10% takes the other 90%.

I forgot to let the Codex VM reach my PXE server through the firewall, so the Proxmox reinstall should have failed. It did not. Codex used the target host as a jump box, staged the PXE stack, armed one-shot PXE, forced the PXE boot with headless-amt, watched the install over KVM, then flipped boot back to disk and disarmed PXE. Both scary and fascinating. Successful /goal run: ~28 minutes, after ~12 hours idle waiting for credentials. Oops. headless-amt is headless Intel AMT control, now being hardened for OSS release.

codex for startup ideas

Last week, we released *Auto-Review* mode in Codex! It is now the default within OpenAI and reduces amount of approvals needed by ~200X. Amazing work from our alignment team. Read the blog at alignment.openai.com/auto-review