# Access and offer evidence

Checked October 8, 2026. These checks are separate from inference measurements.

- [TypeSafe models](https://docs.typesafe.ai/models): directly retrieved official Markdown; versioned `jev-1.13.0`, $0.042 per million input tokens, free output, 64k combined request budget. Versioned IDs may be accepted even though authenticated model discovery lists aliases only.
- [HTTP contract](https://docs.typesafe.ai/api): directly retrieved official Markdown and live OpenAPI; `POST /v1/systemone` with two independent Choice questions. Only inference and model discovery are listed; no public balance endpoint was found in that schema.
- `GET /v1/models` succeeded at 2026-10-08T11:53:52.176Z using the existing key privately. Returned `jev-latest` and `jev-preview`. This verifies discovery, not inference, credit or quality.
- [September 20 promotion](https://x.com/typesafeai/status/2101786280946499671): direct webpage opening failed. The first-party X syndication JSON was retrieved; it says all users started with $5 credit. The record's timestamp is 2026-09-20T21:31:13Z.
- [September 27 notice](https://x.com/typesafeai/status/2104337824292220981): direct webpage opening failed. First-party X syndication JSON was retrieved; it says new signups no longer receive free credits and hopes to restore them. Timestamp: 2026-09-27T22:30:08Z. No later restoration was verified in this check.
- Owner account report, supplied in this task: $5 TypeSafe-direct promotional credit arrived automatically after signup with a Gmail account. Signup date was not supplied. This does not establish a Gmail requirement, recurring benefit or today's new-user eligibility.
- Post-run owner account report on October 8: the dashboard still displayed $5, no billing history was visible, and it said credits were expiring in 9 days. This was not an independent browser observation. No unrounded balance, itemized debit or exact expiry instant was provided; an unchanged display does not demonstrate zero inference cost.
- [Promotional-credit terms](https://typesafe.ai/legal/mca), directly opened: section 8.2 leaves issuance discretionary, permits additional expiry/revocation terms, consumes promotional credit before purchased credit, and forbids multiple accounts for extra promotional credit.

The credit belongs in historical/account experience, outside Active Deals until new-user availability is supported. At the listed rate, $5 / $0.042 × 1,000,000 = approximately 119 million input tokens. This calculation is neither a balance verification nor a measured workload bill.

The official [TypeSafe agent skill](https://github.com/typesafe-ai/skills/blob/main/skills/typesafe-ai/SKILL.md) was read as implementation guidance; no shared agent installation was modified. Raw primary-source retrievals are retained privately under ignored `research/jev-inbox/sources/`; credentials, cookies and account identifiers are excluded.
