back to the doors
/the lab
ch. 00 · an opening
welcometothelab.
currently shipping
  • AI headline optimization engine
  • Landing Page Conversion Auditor
  • cloud-hosted browser automations
ch. 01 · the auditor
01build

Landing Page Conversion Auditor

computer vision audit, on any URL

Screenshots any landing page, runs it through GPT-4o-mini Vision to detect CTAs, forms, headlines, and images, then scores it against a 7-rule CRO rubric and returns an A–F audit.

  • ScreenshotOne
  • GPT-4o-mini Vision
  • TypeScript
  • 7-rule CRO scoring
live · paste a URL below to try it
read the build write-up

screenshot + GPT-4o-mini detection + 7-rule CRO score · ~15s per audit

ch. 02 · other builds
02build

Google Discover Agent

bridges data, editorial, and marketing

Low-code automation that pulls 7-day Google Discover headlines via the Search Console API, processes for quality, and logs results to Airtable.

  • Search Console API
  • Airtable
  • Apps Script
<5 min runtime · 400+ records · resilient retry logic
03build

Data Matching Suite

production web app for messy editorial data

Automates matching editorial data (headlines, writers) with performance metrics — even when formats vary. Fuzzy matching, configurable thresholds, exportable dashboard.

  • Python
  • fuzzy matching
  • CSV/Excel parsing
85%+ match rate · ~70% time reduction · production
04build

Daily Headline Agent

market news → daily stock headlines

Fully automated tool that generates high-performing stock headlines each morning. Pulls market news, formats inputs, GPT-4 generates investor-focused options.

  • GPT-4
  • Apps Script
  • Zapier
  • Sheets
  • Slack
Daily delivery · deduplication built-in · auto-resets
05build

Market Movers Automation

weekly ticker pipeline

Extracts U.S. stock tickers from CapIQ files weekly, updates Google Sheet with date-stamped tabs, syncs to Airtable with Slack pings.

  • Python
  • Apps Script
  • Airtable
  • Slack
1 hr → 5 min weekly · supports 500+ articles/month
ch. 03 · build philosophyexpand
  1. 01

    Ship fast, learn early

    I don't wait for perfect conditions. I start small, test quickly, and let real data reveal what's working — turning ideas into something tangible in hours, not weeks.

  2. 02

    Build for real use, not theoretical use

    Every system has a purpose: does it save time, remove friction, make decisions clearer? If yes, I build it. If not, I rethink. I don't build cool AI things — I build useful systems.

  3. 03

    Data before feelings

    CTR, impressions, session lifts, search trends. Data tells the truth; AI helps me surface patterns faster. Together they guide smart decisions.

  4. 04

    AI is a co-worker, not a shortcut

    AI helps me structure workflows, validate edge cases, process datasets, generate options. But I decide what goes into production. AI accelerates building — it doesn't replace judgment.

  5. 05

    Build to multiply impact

    If I build this, does it increase the team's capacity meaningfully? If yes, I move. Every system I build exists to amplify people, not replace them.

ch. 04 · what's next

currently: shipping a content agent on AWS Bedrock + Kiro.

if you made it this far, I actually want to hear from you.

say hello