notes
Writing, mostly.
Software, AI, poker, money, getting older, and whatever else I am trying to think through without turning it into a content calendar.
I'm trying to write like a person again
16 Jul 2026 · 3 min read
AI is fucking good, but for personal writing I think I miss the messy bit where the opinion forms.
AI code needs better proof
07 May 2026 · 6 min read
AI has made code generation cheap, and everyone is all in, but nobody is thinking about how the review cycle should change.
AI coding has levels
27 Apr 2026 · 6 min read
Vibe coding is one way to use AI. Treating every AI-assisted workflow as the same thing is lazy.
Agents make low-value work feel productive
17 Apr 2026 · 3 min read
Agents can turn a day into a week of code output. That still doesn't mean the work was the right bet.
Hermes Console: a read-only dashboard for Hermes
14 Apr 2026 · 5 min read
Running Hermes locally means checking files, logs, cron output, and CLI state in too many places. Hermes Console is a read-only UI I built to make that easier.
I tried multi-agent orchestration and it was shit
08 Apr 2026 · 5 min read
The honest failure story of building a mission control style multi-agent setup, watching it fall over, and going back to boring workflows that actually work.
One Agent is Enough
21 Mar 2026 · 7 min read
Why most people don't need five agents talking to each other. Skills, crons, playbooks, and memory in a single-agent setup that actually works.
Stop dodging the messy codebase
14 Mar 2026 · 4 min read
The rough codebases everyone avoids are where the real learning and visibility live. A recent dependency upgrade turned into a full cleanup and the team noticed.
I built grep for on-chain events
12 Mar 2026 · 5 min read
I kept doing the same thing at work. So I built the tool I wanted and put it on the internet.
Docs are a performance feature
10 Mar 2026 · 4 min read
In AI-heavy teams, clear documentation isn't overhead. It's infrastructure that makes humans and agents faster.
Mentoring juniors when agents write most of the code
07 Mar 2026 · 4 min read
When agents handle implementation, mentoring shifts from teaching syntax to teaching judgement, debugging, and ownership.
Prompting beat model choice on RoastMaster (until it didn’t)
04 Mar 2026 · 4 min read
I tuned Grok and Sonnet on RoastMaster. Prompt constraints made the biggest difference first, then model personality became the deciding factor.