Building a product that turns tracking into a next move.
Founder-led product design, coaching logic, and iterative delivery with ChatGPT and Codex.
What needed to work.
A tracker can store plenty of data while leaving its user to do all the interpretation. I wanted a product that connected observations, context, trends, and journal reflection to a clear next action—and sounded like a capable coach instead of a report generator.
The ownership I brought.
- Defined the product purpose, coaching voice, daily/weekly/long-term experiences, and the standards for a useful readout.
- Used ChatGPT and Codex to build and iterate the tools, translating domain knowledge and real use into product requirements and implementation feedback.
- Specified a primary tracking source and challenged stale counts, duplicated interpretation, inconsistent navigation, and wording that did not match the intended experience.
- Applied the product to a documented 12-week founder pilot and continued iterating from practical use and quality reviews.
Decisions that shaped the work.
Make the source of truth visible in the behavior.
When totals, dates, and long-term views disagree, polished graphics cannot restore trust. My product requirements prioritize consistent data authority and coherent interpretation across views.
Separate the view from the question it answers.
Today should identify the next useful action. Weekly should explain patterns. Long-term should explain the arc. I pushed against repeating the same commentary across every surface.
Own the judgment when using AI to build.
AI-assisted implementation still needs precise requirements, acceptance criteria, and release discipline. The documented codebase includes coaching modules, adapters, feature flags, route rendering, and contract checks. The July audit also surfaced a source/build mismatch, reinforcing why a successful deployment alone is not proof of release quality.
Different views. Different jobs.
A daily view should help the user act.
A busy teaching day with limited time between commitments.
The context matters. A required class is an instructor obligation, and a useful plan should account for it.
Shape the next action around the schedule instead of adding more activity.
Recreated to illustrate the product approach using sample content. This is not the live tracker or a connection to personal or client data.
What changed.
Created an evolving product with a documented 84-day founder-use record, a JavaScript-based application and coaching codebase, and iterative work across user experience, content, and delivery. Google Sheets, Apps Script code, GitHub, and Vercel appear in the product record. The product continues to evolve through hands-on use and refinement of its data, coaching, and user experience.
I can move between user insight, product direction, AI-assisted implementation, and the standards needed to make the result trustworthy.
Source & outcome context
May 18 product review covering the 12-week founder pilot; July 31 repository handoff; subsequent product direction and AI-building work described by Tai. The walkthrough below uses illustrative content, not a live connection to the tracker.