Use it — when the deliverable is a finished race.
A no-skill control and a skill build, same brief. The skill version is larger, slower to make and visibly more of a product. The control is smaller and already fun.
Read the case
Cases / evidence
Same model, same single-file constraint, different harness. These are not controlled model benchmarks; they are controlled harness comparisons. The question is always: what did the harness change, and what did it catch?
01The evidence table
A case study without the failures is marketing. Ironsight is here even though it was not published: it is the run that produced the product gate.
| Run | Genre | Harness | Size | Gates | Verdict |
|---|---|---|---|---|---|
| Sundown Circuit | City kart | none | 890 KB | — | Strong core, thinner shell |
| Sunset District | City kart | skill + gates | 1.58 MB | 8/8 | Use it |
| Neon Breach | Shooter | none | 304 KB | — | Strong content, stylized |
| Ironsight | Shooter | skill v1 | 1.50 MB | 6/6 | Not enough |
| Black Ice | Shooter | skill v2 | 318 KB | 8/8 | Use it, with a budget |
| Saltline | Shooter | — | 1.42 MB | — | Reference |
Gates are the project's declared acceptance command at the time of the run.Not a controlled model benchmark
02How to read a run
The model is the same across the A/B runs. The harness is the variable. That isolates the harness, not the model. It does not prove that the skill is better than a human, or that 318 KB is better than 1.4 MB.
IT PROVES
The product gate moved the second shooter from a frame with one enemy to a game with four roles, a boss, settings and a save. The run changed because the acceptance command changed.
harness effectIT DOES NOT PROVE
The skill delegates judgement to checks. It still needs a human eye for the art direction, the subject and the final veto. A green gate is a floor, not a ceiling.
human vetoIT NEEDS
A profile captured from another subject ranks style, not quality. Without a project profile the look gate explicitly falls back to tone and emptiness.
calibration03 / Read next
Each case has the brief, the numbers, the screenshots, the gate output and the failures — including the ones the gates missed.
CASE 01 / CITY KART
The control is already a complete arcade racer. The skill build adds the shell, the signature wet road and the product layer. The verdict is conditional, not automatic.
read the case →CASE 02 / SHOOTER
Saltline is the reference. Neon Breach is the no-skill control. Ironsight is the green-but-thin first skill run. Black Ice is the second generation.
read the case →Build one
A container yard, a workshop, a stairwell, a market. The grammar is constant; the mechanics are not. If you build one, open a case study with the failures left in.