Paul Bakaus says he helped turn Impeccable from what he calls a “vibed personal skill” into a design toolkit. He says hundreds of thousands of designers and developers now use it. Three of his techniques show how much of that improvement came from engineering around model behavior instead of polishing a prompt.

The first is “Make them argue · two blind reviewers beat one confident guess”:

For a concrete example, /impeccable critique, Impeccable’s design review command, runs both deterministic checks against your code (to detect stuff like low contrast quickly and decisively) and reviews the design like a human would. But early versions had a major flaw that produced bad critiques: hand the LLM the detector output up front and it skews BOTH ways:

1. detector noisy → condemns a strong page over fixable nits (false alarm)
2. detector silent → rubber-stamps a generic page (meaningless clean bill)

To debias, critique now spawns two sub-agents that never see each other’s work: A = an LLM design director (hierarchy, slop, heuristics), B = the deterministic detector + browser evidence. When the sub-agents finish, the main agent synthesizes (weave, never concatenate) the results from both for a balanced critique. Two blind opinions beat one confident guess.

I agree with the architecture. Review independence matters as much for agents as it does for human teams; the reviewer needs a perspective the maker didn’t already shape.

Bakaus’s second technique is “Force divergence · escape the cluster, don’t chase it with bans”:

Bans just move the model from one region of latent space into another. The escape is sideways, not a longer denylist.

Here are three techniques that produce higher divergence (cheapest but weakest first):

1. Shave the safe picks: make the model name its top 3 fonts, then discard all three. Out-rank yourself.
2. Generate ~50, a blind sub-agent keeps the 5 most distinct. (Lives in Radiant’s codebase, not Impeccable yet.)
3. Seed from outside the model: palette.mjs --from 8f2a starts in a region the model never picks on its own.

This is a better response to repetitive AI aesthetics than accumulating another denylist of fonts and visual effects. The intervention changes the candidate set instead of pleading for a better choice.

The third is “Give them memory · runs that compound instead of restarting”:

Most skill authors treat skills as stateless prompts, but there’s no rule that says a skill can’t have state. Keep in mind that skills can be bundled with scripts that can be run anywhere during the lifecycle of the skill!

Impeccable’s critique and polish commands use this technique quite effectively:

1. /critique writes a snapshot per target (score, P0/P1, markdown + frontmatter)
2. /polish reads the latest snapshot as its fix backlog

One non-obvious detail that makes this work across a team: the snapshot slug comes from the resolved file path, not from how you phrased the request. A teammate pointing at the same file inherits the same memory, and the score trend (24, then 28, then 32) survives across sessions.

This one is immediately practical: save the work on disk, then let the next command pick it up.