v0.2.0 — use cases, treasury funding, expanded front page

Adds a use-case document, corrects the treasury's funding, and expands the repository front page. Live at https://melodic-syrniki-2d566a.netlify.app

Use cases, with the figures behind them

New docs/use-cases.md, published as document 02. Five cases, every number dated and traceable to the document that records how it was measured:

• Before you press buy — $100,000 of TSLA into the wrong pool costs $3,669; $25,000 of NVDA costs $1,364 for the identical 112.709 tokens
• While the market is shut — the reference is frozen for roughly 136 of the week's 168 hours, and an AMC token once ran 35x past the real price on a Saturday
• Around a corporate action — the CrowdStrike x4.0 multiplier, where the naive comparison invents a premium of roughly 340%
• Reading pools against each other — NVDA at 2 bps of basis and 571 bps of spread at the same instant, and the TSLA turnover artifact the ranking rule prevents
• For a protocol — the read signature, honestly gated

Measurements throughout, never advice. The no-advice rule is restated at the top of the document.

The treasury has a funding line

The previous economics claimed the same 10% slice of revenue twice — once for infrastructure, once to seed the treasury — so the treasury had no source of its own, and on the published table it would have taken roughly four years to arm at the lowest volume tier.

Creator revenue now splits 50/50: half buys the token off the market and destroys it, half funds the treasury, permanently rather than only until it arms. Infrastructure is paid from the treasury's line and itemised before profit is struck. The token document now publishes how long the $50,000 threshold actually takes — about ten months at $50k of daily pool volume, two months at $250k, two weeks at $1M — rather than letting "arms at $50,000" read like a near-term milestone.

A repository front page that explains the product

Three new sections in the README: what it is for, with the audiences and their figures; how a reading is built, the full pipeline from three public upstreams through the multiplier, the turnover rule, both formulas and the flag thresholds; and how the project earns, the three engines with their switch-on conditions and the publish-first rule that resolves the conflict of publishing gaps while also trading them.

Also

• Inference is deployment configuration and is not named anywhere in the source, the documentation or the API. The interpret response carries the reading and its grounding only.
• /api/v1/status reports which interpreter settings a deployment has, as booleans, so a misconfiguration is diagnosable without guessing.
• Security headers are set at the edge, so they arrive on every response regardless of which layer answered.
• The copy gate now fails the build on any tool, vendor or model name anywhere in the repository, alongside the existing naming and honesty rules.