stopped writing code
stopped prompting agents
stopped building software factories
stopped engineering outcomes
stopped training autonomous agents
never gonna stop posting.
1
46
It’s never been harder to lead a software engineering team.
Best practices from 60 days ago are already becoming obsolete.
What belongs in CLAUDE.md vs agent skills? Changing.
How do you verify AI-generated code? Changing.
Code reviews? Changing.
CI/CD? Struggling to scale with AI agents.
CVE management? Needs rethinking as AI-powered attack swarms emerge.
Meanwhile, a million vendors are trying to sell you solutions to problems you don’t even know you have.
The hard part isn’t adopting AI anymore.
It’s figuring out how to run an engineering organization when the playbook changes every 60 days.
How is anyone keeping up?
9
10
223
i hate insurance but it’s coming for software engineers / agent builders. doctors have malpractice insurance, developers will have stochastic variance liability and it will cost more than your coding agent sub.
3
5
225
this felt like a game! go poke devin.
blog.google/innovation-and-a… google (neon) and microsoft (MAI) microsoft.ai/news/introducin… is getting serious.. iq/cost pareto will be interesting to watch.
4
131
I’ve played lots of video games, thinking back I don’t think any was worth it. maybe I’m forgetful but I’m skipping the new GTA. When I was playing warcraft and DoTA I could have been building facebook. is it time to mute GTA from the timeline?
Day 1: New AI tool (ADE, bot, coding harness) = magic
Day 3: Hit every edge of the UX
Day 4: Cloning the repo or the app to fix it yourself
Our expectations keep rising. The tools can’t keep up unless UX extensibility becomes a primitive.
The loop continues. Who else is in it? 🔄
#AI #AItools #UX #DevLife #CodingAgents
when are we going to get a continuous variable model size / instance adaptive scaling LLM? i.e. if model sizes are like car gears.. when is the CVT version coming?
2
8
284
giving kirocrew another try… and it’s good?! native extensibility means a much better ux than claude and codex. assistant bots without being locked into cloud grokbot/dots or jumping through hoops with local claude and codex.
1
7
207
model your entire org. everything, especially the communication topology, work delegation and verification. then make it all agentic.
1
4
178
hot take: "what's the right interface for agents" is the wrong question
ask what's the right interface for managing people. that's the actual design problem.
never stop anthropomorphizing.
2
6
209
how to level up? ask yourself - if I had to do what I did today but at 20x scale. what would the world need to look like to make that happen? then go create that world.
1
6
161







