Reviewing plans is a waste of time.
It was useful in the past when planning cognitive load <<< implementation cost. But when implementation is free, just ask the model to bring you 3 complete, end-to-end tested solutions and iterate from there.
And if all of them suck, most likely the model was missing some crucial piece of context.
Show more
To review, or not to review, that's the question.
Reviewing plans is a waste of time.
It was useful in the past when planning cognitive load <<< implementation cost. But when implementation is free, just ask the model to bring you 3 complete, end-to-end tested solutions and iterate from there.
And if all of them suck, most likely the model was missing some crucial piece of context.
Show more
Code reviews today are mostly about gut checking if the complexity is worth the benefits.
To review, or not to review, that's the question.
Code reviews today are mostly about gut checking if the complexity is worth the benefits.
What I want for Xmas
When friends ask me about work:
- Can Codex do XYZ?
- Have you tried asking Codex?
- …
- …
Just ask Codex.
What I want for Xmas
Codex helps me with Slack.
It's so delightful to come back to my desk and see Codex-drafted reply in the Slack app itself.
Codex learns without me having to give it feedback, because it sees what I actually sent and tunes its own behavior.
The prompt I use in the next tweet ↓
Show more
That's a wrap, folks.
When friends ask me about work:
- Can Codex do XYZ?
- Have you tried asking Codex?
- …
- …
Just ask Codex.
Confession: I hated MCP when it came out.
And during these years I changed my mind at least 3 times:
Watch with sound on 🎧
Love how the words animate into existence in the CLI.
If you haven't tried the bidi voice, it's a whole new level of DX, miles ahead of the old text-to-speach setup.
Voice mode is out in Codex CLI 0.155.0.
Enable it with /experimental and then just try /voice.
Codex helps me with Slack.
It's so delightful to come back to my desk and see Codex-drafted reply in the Slack app itself.
Codex learns without me having to give it feedback, because it sees what I actually sent and tunes its own behavior.
The prompt I use in the next tweet ↓
Show more
Confession: I hated MCP when it came out.
And during these years I changed my mind at least 3 times:
How to graph-max with Codex and 5.6 Sol:
1. Draw a graph (literally in any tool, even on paper)
2. Send it to Codex and say "write a code mode script that implements this workflow, run it with <your inputs>"
There's no step 3, it just works.
Seriously it's that simple.
Show more
