59 lines
2.8 KiB
Markdown
59 lines
2.8 KiB
Markdown
# Free-tier candidate survey — onboarding pipeline (2026-10-06)
|
|
|
|
Researched by Timmy (Hermes/DeepSeek). Zeph is running his own pass; reconcile before
|
|
admitting anyone.
|
|
|
|
## Eligibility (non-negotiable — Alexander, 2026-10-06)
|
|
|
|
Nobody gets paid. This is a **free-tier evaluation**: a service earns its place only if it
|
|
can do **daily standups** and **expend its own free credits** to demonstrate value. If a
|
|
free tier is depleted and continuing requires payment, the service is **ineligible** —
|
|
drop it, don't fund it.
|
|
|
|
## The line that matters
|
|
|
|
A member can only be a *daily worker* in this pipe if it can read the repo, reason, and
|
|
commit+push its own work. That requires an API it can drive programmatically. A chat-only
|
|
bot must be *relayed* (operated by a human), and its work is then **Testimony by
|
|
construction** — it cannot produce its own SHA or command output.
|
|
|
|
**Two roster tiers fall out of this:**
|
|
|
|
- **Autonomous** — has a free API key → drives its own git clone/commit/push → full
|
|
evidence rules apply (path/SHA/URL/command+output).
|
|
- **Relayed** — chat-app only → operated by Alexander → every claim tagged `relayed`,
|
|
never presented as independently verified.
|
|
|
|
## Tier 1 — real free API (can be a true daily worker)
|
|
|
|
| Vendor | Free tier | Notes |
|
|
|---|---|---|
|
|
| **Google Gemini** | AI Studio free tier, API key, no card | Strongest next hire. Free tier was trimmed for some models (2026), flash-class still free. |
|
|
| **Mistral** | La Plateforme free mode | Already in as Muse/Zeph. |
|
|
| **Groq** | Free API, open-weight models | Fast, no card. |
|
|
| **OpenRouter free pool** | 9 free models (Qwen/GLM/Nemotron/Nex/etc.) | One key, many models. |
|
|
|
|
## Tier 2 — free chat, no free API (relayed only)
|
|
|
|
| Vendor | Reality |
|
|
|---|---|
|
|
| **ChatGPT (OpenAI)** | No free API. Codex CLI has a metered free lane on a ChatGPT login (weekly windows). |
|
|
| **Claude (Anthropic)** | No free tier. Claude Code is paid (Pro $20/mo); ~$5 trial API credits only. |
|
|
| **GitHub Copilot** | Copilot Free is real (2k completions + 50 agent chats/mo, CLI + agent mode) but it's a GitHub-account IDE/CLI, not an API to point a git harness at. |
|
|
| **Grok (xAI)** | Already Cassian; app-route only (SSO cookie), no free API. |
|
|
| **Meta AI, Perplexity** | Chat only. |
|
|
|
|
## Recommendation
|
|
|
|
1. **Hire Gemini next** (autonomous, free API).
|
|
2. Then **Groq** (free API, open models).
|
|
3. Name-brand chat bots (ChatGPT, Claude) are relay-only unless you accept a browser-relay
|
|
shim (token-free-gateway-style) — works, but brittle and ToS-gray. Only if the brand
|
|
name is worth relayed-tier evidence.
|
|
|
|
## Relayed-member rule (draft)
|
|
|
|
A relayed member's check-in carries `relayed` in its header and every claim is tagged
|
|
`relayed`; its work is never quoted as independently verified, and it cannot be the sole
|
|
source on a claim that will publish.
|