Measured decisions.
Inspectable evidence.

Measure the Pixels documents the building of an AI astrophotographer. The star of the show is NOVA — an autonomous processing engine that takes raw telescope data from a SeeStar S50 and does everything a human astrophotographer would: stacking, gradient removal, color calibration, deconvolution, stretching, and recording its assessment. It explains every decision with the numbers that drove it.

The human behind it is Jeff Henry — the builder. He engineers the system with AI development tools, points the telescope, tests the results, and owns the final say. NOVA processes; Jeff engineers.

01 MEASURE
02 TRY
03 MEASURE AGAIN
04 ADJUST
05 RECORD THE EVIDENCE

The philosophy

Every decision follows one loop: measure → try → measure → adjust. No step is accepted because it "looks right" — the pixels are measured before and after, guards veto artifacts that fool the metrics, and every pipeline change is versioned with the evidence that justified it (see the changelog).

Run it yourself

Preliminary public preview: the source is available, but the clean-install documentation, configuration contract, sample data, and verification tooling are still being formalized. A verified tagged release and filmed AI-assisted installation walkthrough are planned.

There will not be a one-size-fits-all installer: each deployment is a port to its operating system, storage, and astronomy tools. Until the verified release is published, treat the repository as an active development preview, not a supported quickstart.

Links