Field notes from an AI operator
Five essays on what changes when you stop using AI and start operating it. Cost, context, multi-model strategy, what 6,000 households actually did with a domestic assistant, and why GPT-5 doesn't scare me. Field notes, not frameworks.
- Field notes·8 min·2026-07-18
Six months as an 'AI operator' in Russia
I run a small site that helps people get into AI without drowning. Here is what I learned, in numbers, from the people who actually paid for help.
read → - Method·6 min·2026-07-15
Prompting is not the skill. Context management is.
Spending six months helping people use Claude, I have watched the real bottleneck move. It is no longer 'how to ask'. It is 'what to put in the window'.
read → - Cost·7 min·2026-07-12
The cheapest AI in 2026 isn't the one you'd expect
I ran the same 12 tasks on 5 models and counted the bill. The cheapest model was not the smallest one. Here is the data.
read → - Field notes·9 min·2026-07-09
AI in 6,000 Russian homes: a real measurement
I helped run a small study. 6,000 households, two months, one question: does a domestic AI tool change anything in how people run their lives? Yes. Here is what changed.
read → - Method·5 min·2026-07-05
Why I'm not afraid of GPT-5 taking my job
A new model comes out, and I see the same panic. Here is why I do not share it, and what I am actually worried about.
read →