Codex of current studies · Oct 2026
Nine questions I keep coming back to. Each one is a working sketch, not a finished answer. I measure, redraw and revise as I learn, then write up what survives as posts.
Optimal ways to utilize AI for software development
Which habits, tools and review gates actually make a developer faster, and which only feel faster.1
- no posts yet — notes in the notebook
Using AI for a simplified product/requirements workflow
From a fuzzy idea to a testable spec with one lightweight loop instead of a stack of documents.2
- no posts yet — notes in the notebook
What are agents? Really?
A working definition that survives contact with real work: goal, tools, loop, and where a person stays in it.3
- no posts yet — notes in the notebook
What is an AI organizational on-ramp or competency ladder?
The rungs an organization climbs from first experiments to dependable daily use, and how to measure each one.4
- no posts yet — notes in the notebook
High value AI ‘here today’ slam dunks
The jobs that work now, with today's tools, at low risk. No roadmap, no waiting for the next model.5
- no posts yet — notes in the notebook
Best utilizing free or cheap cloud resources
How far free tiers and small bills go before real infrastructure is worth paying for.6
- no posts yet — notes in the notebook
Automating documentation
Keeping docs true as the code changes, by generating and checking them instead of rewriting by hand.7
- no posts yet — notes in the notebook
Understanding and automating/hacking the new SEO/AEO/GEO world
How answer engines and generative search choose what to cite, and what a small site can do about it.8
- no posts yet — notes in the notebook
Playing with hardware (ESP32 and digital ink particularly)
Cheap microcontrollers and e-paper displays as calm, always-on screens for data that matters.9
- no posts yet — notes in the notebook