← adimendelowitz.dev

Adi's Blog

Notes on building instrumented Claude skills, and whatever comes after.

A/B TEST FORENSICS · CASE 01, PART I

Lost Logs, Found Lift, part I

Designing an A/B test before reading its dashboard: what is worth shipping, which comparison, what evidence, which records.

 PAPERS · TMLR 2026

Paying for a Meal That Had Gone Cold

What controlled experiments show about forecast horizon, cross-series information and the cost of attending across series.

 SKILLS · POST 0/6

Skills That Remember

How I stopped treating LLMs like chatbots and started building workflows around them.

 SKILLS · POST 1/6

I Built an AI Reviewer That Remembers Its Last Review

A workflow for separating issues found, patches proposed, and fixes actually accepted.

 SKILLS · POST 2/6

Handoff: A Schema Instead of a Recap

How to preserve decisions, state, and open questions when an AI session ends.

 SKILLS · POST 3/6

Retrospective: Where the Findings Actually Go

Turning AI workflow history into scored decisions and owned actions.
SKILLS · POST 4/6

Token-Aware

Coming soon.

SKILLS · POST 5/6

KB-Search

Coming soon.