// the lab
I do not just use AI. I build with it.
This is where the CS degree earns its keep. Each build is tuned to the business it serves: skills that hold their voice, pipelines that run the busywork, and workflows that audit and draft while I sleep (then wait for me to approve before anything ships).
claude_skills
Client-specific Claude skills
A packaged skill that has read the client style guide, internal-linking rules, product catalog, and past top performers. When it drafts a brief or a meta description, it comes out sounding like them, not like a robot guessing.
- Brand voice and tone baked in, not pasted into a prompt each time
- Internal-linking and keyword rules enforced automatically
- Reusable across the whole content team
make.com
Make.com automation pipelines
Scenarios that watch a spreadsheet or a CMS, run the AI step, route the output for human review, and publish on approval. The repetitive 80 percent runs itself so my hours go to strategy and judgment.
- Trigger, enrich, generate, review, publish
- Human approval gate before anything goes live
- Logged and reversible, so nothing is a black box
audit_engine
Automated audit & meta workflows
A workflow that crawls thousands of URLs, flags the technical issues that matter, and drafts the missing or weak meta descriptions in the client voice. I review the queue and approve in bulk instead of writing each one from scratch.
- Crawl plus prioritized issue list, not a 400-row dump
- Draft metas and title tags ready for one-click approval
- Re-runnable monthly to catch drift
“AI handles the leverage. I handle the judgment. Nothing ships without a human signing off.”
the one rule the whole lab runs on
Curious how any of this works?
Happy to talk shop about building with AI, or trade notes on what works and what does not.