The frontier layer
The AI Lab.
The lab notebook you get to read over our shoulder. New model rollouts, prompt patterns that beat the default, cost benchmarks, what broke, what surprised, what holds.
Latest experiments
LAB-024Claude 4.5 vs GPT-4o on brand-voice consistencyClaude wins on tone stability2026-06-30
LAB-023NotebookLM as a strategist's research deskShipped as a workflow2026-06-23
LAB-022Multi-agent brand audit with n8n orchestrationWorks. Costs $0.42 per audit.2026-06-16
LAB-021GPT-4o Vision for competitor ad teardownsBeats human coders on speed, ties on nuance2026-06-09
LAB-020Gemini 2.5 Pro for long-form category reports1M token window changes the game2026-06-02
Free tier: read-only. Pro tier: full receipts including prompts, configs, and cost breakdowns.
Read the lab. Ship the finding.
Pro unlocks full experiment write-ups, prompt configs, and the underlying workflow used to run each test.