AI Lab · Field notes from real experiments

Here's what happens when
I actually run the thing.

I don't write about AI from a distance. I run experiments, document findings, and publish what I learn, including the parts that didn't work. The goal is to test AI tools rigorously on real design tasks before recommending them to anyone.

Open notebook
Experiments
LAB-01agents · privacyShipped

Local LinkedIn Agent

I built a LinkedIn AI agent. I'm not a developer. It started with a screenshot — and a privacy concern that changed the architecture entirely.

Read the experiment →
privacynothing leaves the machine
modelllama 3.1:8b
chosen viaown benchmark
LAB-02evaluation · toolingShipped
LLM Tester

LLM Tester

A self-built, local-first evaluation environment for testing whether small language models can reliably support real workflows — before committing time, money, or trust to a deployment.

Read the experiment →
purposepick models on evidence
statusin use daily
LAB-03compliance · grcExploring
AI GovLab

AI GovLab

An early-stage AI-driven GRC platform exploring how cybersecurity maturity, compliance workflows, and risk management can operate as a connected system — with AI embedded inside operational workflows, not layered on top.

Read the experiment →
focuscompliance systems
aiinside workflows, not on top
LAB-04hardware · e-inkRunning

Inky Frame Recipes

A Raspberry Pi with an Inky e-ink display showing recipes — built without prior programming experience, using AI as a development partner throughout. Case design in progress.

hardwareraspberry pi + e-ink
casein progress
LAB-05hardware · cyberdeckPlanned

Cyberdeck Lab

Exploring personal computing when the person building it decides what it's for — hardware that is both considered in design and genuinely useful in practice.

thesisuseful, not just pretty