The Data Quality Question a Federal Survey Left Unanswered

A real randomized experiment inside a federal health survey, testing whether a commitment statement improved data quality. The analysis plan was locked and publicly timestamped before any result was seen, and along the way the build caught four independent bugs, all of which would have made a null result look like a finding.

Diagnosing a Segment Failure Hidden by Healthy Aggregates

Win-rate comparison chart showing enterprise vs aggregate customer segments

A synthetic, mechanism-generated B2B sales dataset, 900 companies and 2,600 opportunities across 26 months, where no win rate is an input and every outcome is measured. Disaggregation by segment and open-date cohort surfaces the 25-point Enterprise collapse the blended 20-32% band hid, then competing explanations are tested and the result is tied to an 18-month payback.

Operator Brain Evaluation

Screenshot of the file tree generated by Operator Brain Evaluation from a staged interview.

The public, benchmarked version of Operator Brain. A staged interview writes real files from real answers, and a hard scope boundary is tested and measured, not just claimed.

Decision Clarity

Screenshot of a live coaching transcript from the Decision Clarity skill.

A standalone AI coaching skill rebuilt from a business-specific tool into something anyone can use. The friend test, a reversibility check, and a safety scope tested against five red-team cases.

SQL Instructor

Screenshot of a live SQL coaching session with the SQL Instructor skill.

An AI coaching skill that refuses to hand over the answer. A fixed reasoning pre-flight, a hint ladder that climbs from least to most revealing, and a visual breakdown once the query works.

Executive Assistant

Screenshot of an append-only ledger entry on the Executive Assistant task board.

A reusable AI executive assistant built around one file. State lives outside the chat, updates append rather than overwrite, and “what’s next” always has a real answer.

Make.com Lead Classifier

Screenshot of the Make.com scenario canvas for the lead classification and routing automation.

How a single Make.com scenario replaces manual lead triage. Six boolean signals, a priority-ordered waterfall, and a store-first design so routing failures never cost a lead’s data.

Lead Magnet AI

Diagram of the seven conversation stages in the Lead Magnet AI flow, from capture to insight delivery.

A conversational AI agent that replaces the standard lead magnet form. One live conversation moves a visitor from a stuck decision to a personalized, screenshot-worthy insight. Designed end to end, not yet deployed.

Let's Talk