Eleven themes. One root.
Twelve years of writing distilled into beliefs, evidence, and tensions. Built for me. Readable by agents. The wiki is the source layer; /enter is the playback layer.
Root disposition
Substance over hype
Reduce any hyped category to its substrate, then evaluate. The 8-year through-line from blockchain to anti-customization. The disposition that frames the other eleven.
Themes
Agent-first thesis
Platforms that cannot talk to autonomous agents are already behind. The prompting skill layer migrated; the verdict held.
Voice AI craft
A distinct engineering discipline with its own constraints. The under-share on LinkedIn is deliberate; the case study lives on /lab.
Breadth as differentiation
Breadth differentiates. Depth makes it matter. Humanness-as-depth (EQ, taste, judgment) is the AI-era axis.
PM taste
PM is taste of what to build, what experience, what value. 99% should-we, 1% can-we. Strong opinion about not having strong opinions.
What AI PM requires
LLMs are a primary daily tool. AI replaces the grunt. Taste of what to build is irreducible PM work.
Enterprise AI reality
Roughly 80% of enterprise AI experiments don't reach production. The gap between demo and deployment is wider than vendors admit.
Second brain
Personal context layer for AI sessions. Built this for me, hobbyist scope, Karpathy-inspired. Pattern probably generalizes.
Spec first, taste always
Spec > Sprint. Taste > Execution. Context > Prompt. Lived this since 2021. Named it April 2026.
Career reflection
IC path is legitimate. AI-era amplifies IC leverage. Learning compounds when you study concepts not tools. Help the market flourish.
LinkedIn as instrument
Platform to be gamed productively. Posting IS the thinking. The loop closes when others learn.
Personal projects
Ship it or it doesn't exist. Ideas are cheap; the prototype is the argument. Build, post, teach, learn, build.