People-managing vs. AI-managing
· Jai An
Yishan's post on "AI makes everyone a CEO" has been echoing in my head. I've been shifting from people-management to AI-management as aggressively as I can. Here's what building on his thread looks like from inside that shift.
Some background I'm drawing from:
- Founder/CEO of startups in my twenties with 100+ hires, tens of millions of opex, and experience managing across ICs, managers, directors, and execs.
- Now a founder of a remote-first AI startup with a team of 10–20 (depending on how you count contractors).
---
Yishan's argument, paraphrased
As you move from front-line roles to higher-level roles (founders, execs, investors), more and more time is dedicated to hairy, ambiguous, critical, or strategic decisions. Competent execs hire well and delegate the manageable issues down — so only the hardest things reach the top. Shit rolls up hill. Now that everyone is starting to manage armies of agents doing the routine parts of their jobs: what's left for you? Only the hardest and most critical decisions. The better and more capable your staff, the harder the questions that bubble up, because they take care of everything else. Those become the only duties you have left: thinking REALLY hard about very difficult, ambiguous, strategic decisions. And now it's your entire job. Instead of 10–25% intensity (or less), it's 80–90% intensity.
---
My observations from the people-to-AI shift
AI fatigue is context-management overhead
"AI fatigue" tracks directly with the velocity and difficulty of decisions that AI-accelerated execution unlocks. More throughput means more decisions per hour, and every one of them wants your attention.
My brain OOMs often. Most of the work is modeling the context AIs are running under — any stale upstream assumption derails everything downstream. (OOM = out of memory; the brain version is when you just can't hold another tab open.)
This is very different from the context sharding that happens in people management. There, different departments carry conflicting goals — product wants to ship, compliance wants to vet — and a clash between functions takes change management, 1-on-1s, and real skill to realign.
With AI: find the context gap, patch it, keep moving.
All shit is your fault
This is less "shit rolls up hill" and more "all shit is your fault."
If an AI derails, it's because you fed it the wrong context. There's nobody else to blame, no misaligned incentive, no sandbagging middle manager. Just your model of the work, reflected back at you.
Rest becomes harder
Before AI, the exec/founder brain got rest. Days or weeks of human execution, then an upstream strategic decision. Now execution completes in minutes or hours, and the brain is confronted with decisions at a much higher cadence.
The tradeoff: you can do more with fewer people (less cost, more runway), but you're shifting memory management and context delegation onto fewer human brains. AI can hold clearer context in markdown files, but the person directing it still needs to know which files are current and relevant.
I've had to be more regimented than in a people-CEO role about activities that clear the cache. Your memory management is the AIs' memory management — garbage in, garbage out applies to the whole stack.
RAM is hardwired
The difference between 4GB and 16GB of RAM doesn't matter unless the computer hits OOM. With computers, upgrading RAM is trivial. With brains, the RAM is more hardwired — which means managing what you load into it becomes the whole game.