So excited to play with Mythos/Fable tonight that I can hardly sleep. So I’ve put my agents to work overnight. Can’t wait to see what they’ve cooked up by morning.
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
A joke, but only just.
We’ve been outsourcing parts of cognition to tools for centuries. Writing did it for memory. Search did it for recall. AI is bringing that process closer to higher-order thinking, which is why it feels so personal.
Like so many of my generation, I watched James van der Beek as Dawson on Dawson's Creek growing up. His untimely passing has me feeling not just old, but mortal.
no way this isn’t killing manual customer research
Format connects to Slack, Intercom, Zendesk, Email, etc
then it sends you a 15-minute weekly report with friction, churn risk, and product requests
you can read it, watch it, or listen like a podcast.
here’s how it works:
@aryanlabde If you're an rockstar engineer, hire the killer marketer.
If you're a killer marketer, hire the rockstar engineer.
It takes a well-rounded team to build a business.
@kimmonismus What I like about it compared to Comet is that the browsing agent is *my* ChatGPT - i.e. it knows me and has that additional context. That gives me more trust that it can do things on my behalf.
I had a rare chance to test two coding agents on the exact same hard prod issue. Same prompt, same context. In this case, Claude Code beat out Codex with a clearer root cause and actionable fix. Controlled comparisons are super valuable. Less vibes, more evidence. #AI
Been debugging a “works on my machine / dies in the cloud” saga so long I’m pricing couriers to ship my Mac Mini to us-central1 and call it “the cloud.”
AI coding tools aren’t just great at writing code, they’re beasts at reviewing it.
Code reviews used to mean waiting days for someone to find time, then skimming through 500 lines and saying “LGTM 👍”.
Now I assign Copilot, Claude Code, Codex & Bugbot to every PR.
They all catch different stuff, and keep going until it’s clean.
What I thought was a 2-commit fix turned into 18 commits of polish.
AI reviewers don’t replace humans devs, they amplify us.