04 / What the work changes
A learner does not only receive an explanation. They form an expectation, encounter a spatial or semantic result, and inspect the evidence that supports or resists it.
03Interactive instruments
Prototyping · 2026
Al.Floys
Interactive environments for understanding modern AI.
Executable learning experiences for models, embeddings, attention, and the concepts beneath contemporary AI systems.
Explore Al.FloysInstrument 01 · sampling temperature01 · commit to a prediction
Turn the temperature from 0.10 to 2.00. What happens to the order of the six candidates?
A language model scores every possible next token, divides all of those scores by one number, and then samples. That number is the temperature. The scores below are fixed; only the divisor moves.
0.10 decisive2.00 open
- structure
- memory
- policy
- noise
- habit
- chance
Entropy 1.49 bitsTop two gap 0.393Order unchanged at every temperature
01 / The question
What becomes learnable when an abstract AI concept is turned into an instrument that can be predicted, inspected, and tested?
02 / What exists now
A live browser-based learning environment with focused modes for AI concepts, including an embeddings experience organized around prediction, inspection, and evidence.
03 / One real artifact
Embeddings as an evidence loop
Current interaction grammar
- 01Predict
- 02Look
- 03Evidence
05 / Uri’s role
Uri designs and builds the product, its learning interactions, technical architecture, and visual instruments.
06 / Status & provenance
Active public prototype. The live product route remains separate from this site’s study surfaces.
Last true ·
07 / Onward relations
Contrast
OntogonyAl.Floys makes model relations inspectable for learning; Ontogony makes consequential system decisions inspectable for institutional use.
Foundation
The Transductive UniverseTTU develops the philosophical account of relation and becoming that Al.Floys turns into manipulable instruments.