Skip to content
All findings

Browse by type

Every piece organized by what it does for you.

Recipe4

Tools you can use today

Experiment6

Data and results from testing

Mechanism9

How and why things work

The Cheap Half of the Loop

7 minJun 27, 2026

Coding pulled ahead because its feedback is cheap to check. That is one half of the loop. The same training trains out the other half, and the part that chooses the frame is the last thing to get cheap.

Why AI Defaults to Generic

4 minJun 22, 2026

Every prompt technique is one move: make the default path expensive enough that the model leaves it. Specificity is the largest measured version.

The "It Depends" Problem

Three AIs, No Source, the Same Answer

6 minMay 25, 2026

Same model, same prompt. The source you paste, not the prompt you write, decides whether the numbers are real.

The Fabrication Problem

How AI Makes You More Wrong With More Analysis

6 minApr 26, 2026

Five hours of analysis, increasingly sharp, increasingly wrong. The frame amplifies. The reframe is the fix, not the analysis.

The Model Mechanics

Stop Calling It Hallucination

7 minApr 25, 2026

Hallucination is six or more distinct failure modes. Different mechanisms. Different solutions. Name the type first.

The Model Mechanics

Why Experts Miss What Beginners Catch

6 minApr 25, 2026

Generation builds the mental model that makes evaluation possible. Without it, evaluation collapses to surface features.

The Evaluation Problem

Stop Polishing, Start Switching

4 minApr 16, 2026

The ceiling is per generation mode. Switch modes to access territory that iteration can't reach.

The "What You Think Works" Problem

Four Layers Produce Every AI Output

4 minApr 3, 2026

Four layers produce every AI output. The company's system. Your system. Your prompt. The model. The model is the only one with a name.

The Model Mechanics

Why AI Can't Check Its Own Work

4 minMar 23, 2026

The agent reported clean. The output was wrong. Same process generating and evaluating.

The Fabrication Problem

Arc1

Journey across multiple experiments

From practice2

Findings from building

Practice4

Self-experiments you can run in five minutes