← Nisarg GurjarMCEmail me
INTERACTIVE · TRY IT LIVE

Learning Dashboard. A curriculum for auditing code you didn’t write.

Status
Feature-complete · hosted on this site
Type
Personal · my own learning system
Stack
One self-contained HTML file · vanilla JS
Scale
5 milestones × 3 modules · token-level code reader
Source
github.com/ngurjar513/learning-dashboard
Year
2026
Live
Open it · runs on this site ↗
Runs right herethe artifact is the proof, so open it
The Learning Dashboard: milestone navigation, core concept modules, and copyable auditing rules in a dark interface
Milestone 1 of 5. Don't take the screenshot's word for it; open the real thing above.
The problem

If AI writes your code, your job becomes auditing. And nobody teaches that. I needed a way to systematically learn to read structure, spot bloat, and interrogate code I didn’t write. So I had the AI build the curriculum that teaches me to check the AI.

What I shipped
  • ◆Five milestones. Structural anatomy, data contracts, control flow and edge cases, architecture and bloat detection, test-driven orchestration.
  • ◆A clickable code-anatomy reader. Every token in a TypeScript sample explains itself.
  • ◆Copyable audit rules and adversarial assessment prompts. Tools I actually use in real sessions.
  • ◆Zero infrastructure. One HTML file, progress in localStorage, no build, no server.
How it works
Learn → audit → verify
  1. 1Work a milestone’s three concept modules
  2. 2Practice on the interactive code reader
  3. 3Take the audit prompts into a real session
The hard part
Honesty about the audience: it teaches me. An operator who audits code rather than writes it, which forced every explanation to earn its jargon. The assessment prompts are layered adversarially, so the AI being audited is instructed to make realistic mistakes, and the fail-safe protocol keeps hints from becoming answers.
Proof
  • ✓Open it The entire artifact is one click away, running on this site.
Milestone 3 · control flowIN PROGRESS
Code anatomy · 41 tokensCLICKABLE
Audit prompt · layeredCOPY
Where it stands

Feature-complete and in use. It’s also the clearest statement of how I work: I don’t just trust the code; I train myself to judge it.

I wrote the curriculum and audited every module against it: a course on auditing AI, built by auditing AI. Claude Code implemented it. How I work →

Want something like this built?

I take on projects where shipping something real is the whole point: tested, verified, in production.

Email me