Claude Code and Codex aren't competitors โ they're collaborators. I've been using the Codex plugin inside Claude Code, and the magic is in the handoff: "have
codex review your findings" produces genuinely better output than either alone. The whole is greater than the sum of its parts.
Install the plugin: https://t.co/aJzEFRNnl5
lazygit to watch changes in the codebase is a low key way of making sure the coding agents aren't going off the rails and touching files they shouldn't be.
here is the AGENTS.md:
---
## Claude Compatibility
This repo intentionally keeps Claude-oriented files as the primary agent context. Agents working here should use the local Claude files directly instead of requiring duplicate Codex-specific copies.
- Read `CLAUDE.md` first for the main high-signal repo rules.
- Read `.claude/memory/MEMORY.md` for repo-local memory when it is relevant to the task.
- Use `docs/agent/README.md` and linked docs for expanded reference material.
- Treat `.claude/skills/` as the canonical repo skill directory. When a task matches a local Claude skill, open that skill and follow it even if there is no mirrored skill under `~/.codex/skills/`.
- Skill entrypoints may be named either `SKILL.md` or `skill.md`; check both when inspecting `.claude/skills/`.
- Treat `.claude/commands/` as supplemental workflow documentation/macros when a command name or workflow is referenced.
Claude outage got you down? Configured Codex to honor my existing Claude repo context without duplicating everything (Plus I really like the Codex 5.4 model output).
What worked:
- keep CLAUDE.md as the main instruction file
- add a small AGENTS.md shim that tells Codex to read CLAUDE.md, .claude/memory/MEMORY.md, docs/agent/README.md, and .claude/skills/
- treat .claude/skills as the canonical repo skill directory
- support both SKILL.md and skill.md
Important detail: donโt symlink all of ~/.codex/skills to ~/.claude/skills because Codex keeps its own built-in .system skills there. If you want sharing, symlink individual skill folders instead.
Net result: one repo can work well with both Claude Code and Codex, with Claude files as the source of truth.
AGENTS.md below
@AskWorkspace The gmail web client has a major bug when replying to an email that has inline images. If a non-gmail client sends you an email with inline images, and then you click reply, then gmail converts these into attachments. Iโve been told there is an open Bug ID. ETA?
@gmail@abgoldberg Here is a link to a bunch of other people complaining about the same issue. https://t.co/pqzJp7PVoZ
Any assistance is very much appreciated!
@gmail@abgoldberg A similar bug that has been happening for a long time is when the gmail web client saves a draft of a reply that contains inline images. When you open up the draft later, the images are no longer inline. We have seen this bug for a long time. It seems like it could be related.
@gmail@abgoldberg We are seeing the same issue across multiple customers. The gmail web client takes inline images that are in a reply chain and then makes them into regular non-inline attachments. This is relatively new bug/behavior that started about a week or so ago.
I came from hospital with a remodulated immunosuppressive therapy, but after a week the blood virus concentration skyrocketed reaching the double of the previous week value (765k copies/ml vs 359 copies/ml) and this is seriously bad, because it means my organism doesn't react to the standard therapy (which is actually an anti-therapy).
Since this is a quite rare condition (as most part of patients react well to the immunodepressant reduction), it means mine is a severe pathology seriously putting at risk the new kidney. I read tens of papers from everywhere in the world and apparently there's no effective anti-viral drugs agains this virus (Polyomavirus BK) but there are potential cocktails along with IVIGs, intravenous immunoglobulines.
I suspect I need help with urgency from anyone could know very specific information about experimental treatments for this virus in a post-transplant condition that can avoid the graft loss. Obviously, being in Italy, the quickest help can come from here or from the closest countries, but contacts are important too across hospitals and clinics around the world, so that's why this is a tweet in English (on top of the fact that the account has always posting in English).
Thanks in advance for your help and support as usual.
How a book written in 1910 could teach you calculus better than several books of today
[Calculus Made Easy, by Silvanus P. Thompson, 1910 - full text pdf: https://t.co/W4gspJGhHK or with the table of contents: https://t.co/55dkZzX4XV]
Last year, I had a conversation that changed my life.
It caused me to upend everything and move across the country.
The lesson from it may change yours:
Oof. Iโve been an entrepreneur a lot longer than Iโve been a VC, and I never understand VCs who shit on founders. Whatโs the upside? A few โlikes?โ I believe in the future, the misfits, the bold. Those who get in the arena. Iโd never take money from a VC who dunks on a founder.