I think the "difficulty" of software engineering is essentially constant no matter what abstraction level you move to, because human cognition adapts to new tools until it can fully utilize itself. Tools are only affordances, not a magic wand that makes work disappear. Great software engineering was immensely challenging before. It is still immensely challenging now, despite very different workflows.

America is banning AI in schools. China is using AI to create geniuses. Introducing Aristotle: The AI tutor that solves America’s broken education system. heyaristotle.com

A funny thing is that LLMs are not particularly good building agentic systems. For the moment, you need human expertise to do it well.

I checked in with my friends at Meta to see how AI is reshaping the PM team. What I found is a story I usually only hear at hot startups: endless meetings, reviews, and docs giving way to PMs coding, less process, and far more opinion than I saw in my four years there. So I asked my good friend @jagjitchawla to discuss on the the Skip podcast. A ton of great insights: ◆ VPs used to spend hours reviewing status. Now an AI agent he wrote reads every code change, email, and doc overnight and hands him a punch list before the team logs on ◆ The products got better too — agents triage tens of thousands of bug reports a day and ship the fixes ◆ AI broke things too — site incidents climbed, so they built safeguard agents to vet changes before they ship ◆ Ideas got cheap, so judgment is the bottleneck ◆ The generalists are back — his strongest PMs are economists and physicists, not ex-engineers ◆ It's all homegrown — no vendor, no mandate, built by the teams themselves, some in a single evening ◆ The rollout was deliberate — Meta canceled a full week of meetings, then named "AI captains" by function ◆ Zero-to-one still starts at the whiteboard ◆ What AI doesn't replace: people fall in love with people, not projects You can replicate most of this with little budget and not much effort. Jagjit and I go deep on the latest episode of The Skip: Write-up on Skip: open.substack.com/pub/theski… piped.video/5y3kx8TTsaI

the thing about loops is that you need to build trust first. a low trust loop is just a waste of tokens. and if you don't apply engineering taste, the slop really compounds. but you can become a quality loop enjoyer too with pstack as a starting point: github.com/cursor/plugins/tr… use these skills to write automations that automatically triage, repro, fix, and verify their own work while you touch grass. use them to prototype new ideas and refactor your code and trust that they'll do things the right way

Introducing Roughdraft! A new open source project designed to make collaboration with agents better. The idea is to bring commenting and suggested changes to markdown (e.g. plan docs) in a nice interface. Free, local, etc. 👉 roughdraft.md 👈