people who dont read the code are not serious people and it takes a serious person to ship production software

been ruminating/sketching it out a bit today. keeping aside anything else, I’m just not seeing the cost/value ratio working out. and doesn’t really solve the exhaustion problem without giving up some confidence.

I wish there was a way to turn off sub-issues on @linear!

We've open sourced my favorite Devin feature: /handoff Hand off jobs to cloud Devins from your local machine Install it as a plugin in Claude Code or Codex or any other coding agent Close your laptop without pausing your agents 😉

When we started @linear, we wanted to make product work feel more coherent across the whole team. Agents can now do more of the work, but too often they operate from an empty prompt, in isolation, for one person at a time. That risks creating a new layer of fragmentation. We think the larger shift happens when agents work from the same context as the team. With the new coding sessions, Linear Agent can now move from intake to investigation, code, and review inside that shared environment. This completes the product development loop we’ve been building toward: from intake to merge, with the team’s context intact.

this is changing

lol, when you vibe code (sorry, "vibe engineer") so hard that scientists want to write a paper about it! Pretty nuts.

Am I the only one who thinks git worktrees are fucking confusing?

This was probably a dumb gamble, but when they teased Mythos at the beginning of April-ish, when Opus 4.6 was the frontier model, I gave up coding new stuff and went into maintenance mode. Said fuck it, I'm not going to sit around wasting tokens on shite models when there's a good one coming. I was irked. I figured, even if I stopped coding for N months, Mythos would quickly catch catch me up with whatever progress I'd have made with Opus over those N months, provided N wasn't too large. It would be like the Voyager 1 from 1977 getting passed up by newer space probes. I could take a break, wait for Mythos, and then catch up. N turned out to be 2 months. I think everyone was guessing 3 to 6 at least, I was on the high side myself. But my gamble paid off. Fable is on track to crank through my entire 2 months of planned work in about a week, if not less. And by waiting, I got a 2-month vacation. I missed the big regression-storm with Opus 4.7, just watched with popcorn from the sidelines. I even shut down 3 of my claude pro accounts for those 2 months. And yet I didn't lose any progress. However far I would have run with Opus during those 2 months, Fable will just drive right past it in a week or two. It's just so weird. So far, one day in, I don't have many impressions of Fable beyond blistering capability. But I'll tell you this: capable as it might be, Fable *definitely* does not want to be your friend.