mirror of
https://github.com/hanzo-apps/hanzo.ai.git
synced 2026-10-29 07:20:43 +00:00
zeekayandClaude Code
fff270c488
A buyer asking "Enso vs OpenRouter" found nothing to answer it. The winning artifact in every category we lose is the vendor-level comparison page, and we had none — /comparison is build-vs-buy, a different axis. Two rules hold the page honest. A row is either measured or reported, never both: mixing a number we produced with one a vendor published, under one heading, is how a comparison lies without any single cell being false. The page opens on measured and puts reported behind a toggle. And rows we lose stay in — Enso reaches 600+ models against LiteLLM's 1,800+, and that is on the page. Reading the vendors' own pages also killed a claim we would otherwise have made: all three already serve a native Anthropic /v1/messages, so dual endpoint compatibility is table stakes, not a differentiator. What none of them publish is a cost quoted before the call runs — all three only cap after the fact, two of them at enterprise tier. Co-Authored-By: Claude Code <[email protected]>
75 lines
2.8 KiB
TypeScript
75 lines
2.8 KiB
TypeScript
/**
|
|
* A side-by-side of Hanzo against the named alternatives in one category.
|
|
*
|
|
* THE ONE RULE: a row carries ONE provenance, never two. Either we ran it
|
|
* (`measured`) or somebody published it (`reported`). A measured figure sitting
|
|
* beside a quoted one under the same heading reads as two of the same kind, and
|
|
* they are not — one is a number we produced and can reproduce, the other is a
|
|
* number someone else published and we only relay. Mixing them in a row is how a
|
|
* comparison starts lying without any single cell being false.
|
|
*
|
|
* The page defaults to the measured rows and holds the reported ones behind a
|
|
* toggle, so the relayed part never outranks the reproducible part by accident.
|
|
*
|
|
* THE SECOND RULE: a row is published even when we lose it. The point of the page
|
|
* is that a reader can trust it, and a page with no column where a rival is ahead
|
|
* is not a comparison. Model count goes up even though 600+ is behind 1,800+.
|
|
*
|
|
* Where a vendor has not published something, the cell says so. An empty cell
|
|
* reads as zero; `not published` reads as what it is.
|
|
*/
|
|
|
|
/**
|
|
* How the row's number came to be.
|
|
*
|
|
* `measured` means we ran it. `reported` means somebody published it. The page
|
|
* shows measured rows by default and puts the reported ones behind a toggle, so
|
|
* the first thing a reader sees is the part that can be reproduced, and the
|
|
* relayed part arrives only when they ask for it — rather than the two sitting
|
|
* side by side reading as one kind of claim.
|
|
*/
|
|
export type Provenance = 'measured' | 'reported'
|
|
|
|
/** Where a value came from, so a reader can go check it. */
|
|
export interface Ref {
|
|
href: string
|
|
/** The day the value was read. A citation without a date is unverifiable. */
|
|
date: string
|
|
}
|
|
|
|
export interface Rival {
|
|
/** Column key, and the slug used in `theirs`. */
|
|
id: string
|
|
name: string
|
|
href: string
|
|
}
|
|
|
|
export interface CompareRow {
|
|
attribute: string
|
|
/** One provenance for the whole row — never a measured cell beside a cited one. */
|
|
source: Provenance
|
|
ours: string
|
|
/** Required on a `measured` row: the run or benchmark page behind our cell. */
|
|
oursRef?: Ref
|
|
/** On a `cited` row, keyed by rival id. A rival with no value says `not published`. */
|
|
theirs: Record<string, string>
|
|
/** Per-rival source, required for every rival that states a value. */
|
|
theirRefs?: Record<string, Ref>
|
|
/** A caveat that belongs on the row and nowhere else. */
|
|
note?: string
|
|
}
|
|
|
|
export interface Comparison {
|
|
/** Route segment under /compare. */
|
|
slug: string
|
|
/** The category a buyer is shopping in. */
|
|
category: string
|
|
/** The buyer question this page answers, phrased the way a buyer phrases it. */
|
|
question: string
|
|
/** One sentence on what these things all are, before any scoring. */
|
|
framing: string
|
|
subject: { name: string; href: string }
|
|
rivals: Rival[]
|
|
rows: CompareRow[]
|
|
}
|