@bcherny Hey Boris - focus is something I've been wanting to use for a while, but upon trying it with the latest it doesn't seem the command is available. This is on v2.1.111 which seems to be the latest - could you share if it's on a yet unreleased one?
@andrewchen As someone who has been coding for 20+ years - I'm down!
The hard part never really was coding, it was knowing what to build and how to build it.
I'll still miss spending days in a codebase coding though, it's to me very fun and almost therapeutic.
Looking in the new @AnthropicAI Managed Agents - overall excited about the direction & architecture, makes it so much faster to ship agentic work!
There are concerns about centralizing orchestration at this level but Anthropic's commitment to AI safety makes me feel better.
Finally wrapped up my blog series on AI code gen with a very practical use case - auto-generating PRs from Jira issues!
Read on https://t.co/k5sRoAohNE
@gitconnected@TreyHuffine
@konnydev Europe has the rules and regulations to extract money from Claude and Deepseek.
But more seriously, @MistralAI is doing some decent stuff although they're very far behind other frontier labs, but their strength is probably on open-weight and being EU-native.
Very true, but that also brings a problem - if intelligence ~= compute then how do smaller companies with lower funding compete with behemoths who can give unlimited compute to their engineers?
@SergioRocks True story. Seen that across the industry.
But I'll go further, even director+ levels are being flattened, heck I see sr dirs and VPs coding and pushing PRs now - that did NOT happen even a year ago!
@craigzLiszt Hard disagree on that. Have you ever tried building features in a gigantic codebase? Even top of the line frontier models seem to get lost now in my experience.
Will they get better? Probably. Will I ever trust them to the point of going to prod with zero human review? Def not.
Struggling with the proliferation of skills nowadays - tools like https://t.co/giLJGQxVy7 are great but then you have multiple skills doing the same thing and little in the way of evaluation to know which works best.
Really need an industry standard for skill evals I feel...
Wrote my second article on agentic code generation which was published yesterday with the good folks of @gitconnected@TreyHuffine
You can find it under https://t.co/nDZLnfEgxj and hopefully learn a thing or 2!