AI reliability
Evaluation, failure modes, and the small checks that make an intelligent system easier to understand.
Field notes / 01
A growing journal from an independent experimental engineering workshop. Notes on making, testing, and reducing ideas until they can hold their shape.
Release conditionEntries appear when there is a real artifact, a useful question, or a result that can be described plainly.
The journal will move across a few recurring areas. The labels are a map for future notes, not a catalogue of finished products.
Evaluation, failure modes, and the small checks that make an intelligent system easier to understand.
Bounded investigation, adversarial thinking, and evidence that keeps a technical claim honest.
Local tools and playful interfaces where craft, responsiveness, and a good experiment meet.
Practical notes on the systems, techniques, and choices that make work possible.
A note can start with a question, a broken thing, or an instrument worth trying. It earns a place here by carrying enough evidence to be useful to somebody else.
Records of what was made, tested, and learned across the four lanes.
How a case mismatch exposed the query behind Intigriti's September challenge, and why the XSS hypothesis did not survive its control.
A wordless seven-shot film, from a procedural Blender conservatory to original sound, a measured cloud desktop, and the limits of its verification.
For a focused conversation about the work, send a note with the context and the question you want to examine.