mirror of
https://github.com/hanzo-apps/hanzo.ai.git
synced 2026-10-29 07:20:43 +00:00
zeekayandHanzo Dev
d58d99ae86
/benchmarks now reads as one story in five products: Enso (intelligence per dollar), Kai (decisions), Zen 6 (speed on Hanzo hardware), the agent harness and the agent platform. Each section opens with one headline and one number, then its chart. Methods and the searchable shelf follow. It publishes wins and the product's own measurements, each a recorded run: - Kai leads Jev on 11 suites, listed by name. The Jailbreak suite and the typed HarmBench constants are gone, and so is "Decision Index v2", whose 12th row was not a HarmBench run. - No jailbreak-board figure is published. The Jev run on its unseen rows (hanzo-inc/kai 1671eca) has Jev ahead. - Enso shows the frontiers it holds: GPQA-Diamond and LiveCodeBench v6. - Zen 6 shows its 2026-09-23 readings. - The platform section reads the cloud bench (lib/data/bench.json) and the runtime passes. - The ledger keeps wins and standalone measurements; the leaderboard lists the suites Kai leads on the metric drawn. Monochrome: Hanzo is in the brightest ink and the field in the mid grey; standing is a glyph and a word. Icons come from gui's lucide set, and e2e/gates/glyphs.spec.ts fails on an icon at 0×0, an <svg> with no shapes, an icon outside 12–24 px, or an arrow typed as a character. Both schemes are covered at 375 and 1440 px, as is the contrast gate. Co-authored-by: Hanzo Dev <[email protected]>
8 lines
380 B
TypeScript
8 lines
380 B
TypeScript
/**
|
|
* The five products /benchmarks tells its story through, in its order:
|
|
* intelligence, decisions, speed, then what the agent runs on. No imports, so the
|
|
* browser's shelf and the build's ledger (lib/results) read the same list.
|
|
*/
|
|
export const PRODUCTS = ['Enso', 'Kai', 'Zen 6', 'Agent harness', 'Agent platform'] as const
|
|
export type Product = (typeof PRODUCTS)[number]
|