Hello, BuildHub
Why a non-engineer is starting a daily build log, and what to expect from the rest of the posts.
This is the first post on BuildHub, and it exists for one reason: to force a cadence.
Why a daily log
Building in private is easy. Nobody sees the weeks where nothing works. Nobody sees the branch you deleted at 2am. The problem with easy is that it makes it very simple to quit, and pretend you never tried.
A daily log fixes that. If I have to write about what I did, I have to actually do something.
What this will be
Every post is one day of building a real automation system. The system itself is a trading research pipeline — cron jobs, data processing, a Python worker, LLM-driven decision support. The specifics of what it does are less interesting than the specifics of how it gets built:
- What broke today, and how long it took to find the cause
- What I asked Claude to write, and what I had to push back on
- The commands that almost worked, and the ones that actually did
- The design choices that looked right in the morning and wrong by the afternoon
What this won't be
- Trading advice. You will not find ticker symbols, entry prices, or position sizes anywhere on this site. That's not the story.
- Courses, guides, or "how to build X with AI in 24 hours." Real systems don't ship in 24 hours.
- Perfection theater. If a post reads like I knew what I was doing the whole time, I failed to write it honestly.
Key takeaway
Showing up daily with something honest is the only promise I'm making. The quality of the work will improve because the quality of the work has to improve — it's being read.
See you tomorrow.
// related
- · 11 min read
From Nine DRY RUN Failures to Go-Live: Joseph's T+2 Triple Lock, 458 Tests, 18 Crons
Joseph, an automated investment execution agent, entered live observation on 2026-04-20. This is the full technical write-up — nine failure modes, the T+2 triple lock, Shioaji pitfalls, and every cron in the schedule.
#joseph#claude-code#dry-run#triple-lock#shioaji#automation#trading-system#series-3 - · 10 min read
Seven Principles I Learned Working with Claude: Spec-Driven, Boundary-First, Decisions That Stick
Three months of pair-building Joseph with Claude Code gave me seven principles that let the AI actually run the development loop on its own — not babysitting-style collaboration where you hover over every step.
#ai-collaboration#claude-code#spec-driven#methodology#joseph#series-2 - · 11 min read
I Killed Three Projects Before Joseph Survived: A Non-Engineer's Three-Month AI Build Log
Retail worker in Hualien, not an engineer, four projects, three killed. From CAIOS to Joseph — an investment agent that went live on 2026-04-20. Honest post-mortem, not a success story.
#ai-collaboration#claude-code#non-engineer#project-kill#joseph#series-1