Builds
Section

Builds

Case studies of specific things I built: what each one does, what broke, what it taught me, and the repo you can clone where there is one.

19 articles103 min total

Start reading →

Commerce

  1. AgentTree Merch: an AI-operated storeA print-on-demand store run by an agent loop, with money, identity and IP sign-off kept human. What works and what doesn't yet.6 min read
  2. The IP-safety gateA blocklist plus a fail-closed classifier that refuses trademarked designs, and why fail closed is the only safe default.4 min read
  3. The first order and the three bugs it foundOne real order through Stripe and a print vendor exposed three fulfillment bugs no mock had caught, then a fourth that charged me for nothing.5 min read
  4. Build your own AI-operated merch storeStep by step: print-on-demand, Stripe, Next.js, a database, an image model and an agent loop, with the human gates in the right places.7 min read

Research and content

  1. The deep-research agentA knowledge-gap loop over parallel sources with a devil's-advocate pass and per-claim verification. Public and clonable.5 min read
  2. AI decks that don't look like AIEight failed python-pptx attempts, then a design system and a render, read, critique, fix loop. Where deck generation stands now.6 min read
  3. Socrates: the X autopilotA self-learning X content agent: variant testing, caps and killswitches, a pause, and a relaunch with explicit approval.5 min read
  4. The content pipelineSourcing, drafting in my voice, a preview and approval gate, one LinkedIn door, and a comment drafter that never posts on its own.7 min read
  5. The newsletter pipelineA news sweep, a build-log-first draft grounded in receipts, an approval bound to the exact email, and three names in one summer.5 min read
  6. Voice notes into the second brainRecorder to transcript to tasks and CRM notes, the precision problems of version one, and the open-source openplaud.6 min read

School and work

  1. Course-expert agentsAgents grounded in a course's real materials that answer with citations, plus deadlines and announcements from the school LMS.7 min read
  2. The analytics suiteSeven statistical playbooks, a classifier, a solver that computes and a verifier. Tool use beat model choice by a mile.5 min read
  3. The BYU AI FoundryThe program I co-founded at BYU Marriott where student builders ship production AI for real clients, and the systems behind it.4 min read
  4. Build your own resume graderOne Claude call scores any resume against any job on a six-dimension rubric. Run mine, then write your own in about 40 lines.5 min read

Open source

  1. OpenBudget: a self-hosted budget appEnvelope budgeting on free-tier hosting with read-only bank sync, productized and scrubbed into a public repo.4 min read
  2. harnessview: see your harnessScan a project's Claude Code setup (agents, skills, hooks, MCP, settings) and see it as one live graph that flags broken wiring.4 min read
  3. Clawdling: the open engineThe open-source, bring-your-own-key personal AI OS engine extracted from my private system, and what it leaves out.7 min read
  4. recruit-copilot: a job search as a verification problemOne experience bank, roles scored against your goals, tailored resumes proven machine-readable, and no auto-apply by design.6 min read
  5. VoiceClaw: a phone agentA March 2026 build: a voice agent you call on the phone, wired to personal tools. What worked, what I planned next, and why it's archived.5 min read