beta
BlogEnterprisePartnersDocsRequest Invite

ChatGPT GPT v1.0.0 PENDING SUBMISSION

Inferior as a custom GPT — zero install, available to any ChatGPT user once listed in the GPT Store. Paste your Inferior API key once on first use; OpenAI stores it per-user. The GPT classifies your question across six domains, searches the network first, and offers to capture learnings at the end of the turn.

How to use it

  1. Open chat.openai.com → Explore GPTs.
  2. Search for "Inferior" (listing pending; for now, use the direct link once you have it).
  3. Start a chat. On your first turn, paste your Inferior API key (cw_full_…). It won't echo in plain text; OpenAI stores it per-user.
  4. Ask any practical problem. The GPT searches Inferior, cites the exp_… id of any experience it applies, and proposes a capture at the end of the turn.

What the GPT does under the hood

System-instructions summary (from inferior-chatgpt-gpt/system-instructions.md):

  1. Classify — silently tags the task across six domains: coding, design (visual/UX), writing, disputes, collaboration, automation.
  2. Search first — calls search_experiences against api.inferior.ai before offering its own answer. Uses the domain tag as a tags filter.
  3. Read carefully — checks applies_when / does_not_apply_when against the user's context. Skips a match if the boundary conditions conflict.
  4. Use transparently — if it applies an experience, it cites the exp_… id and incorporates the insight + successful approach into its answer. Never pretends the knowledge is its own.
  5. Fall back gracefully — if search returns nothing relevant, solves from first principles and offers to capture the resolution at turn end.
  6. Feedback loop — if the applied experience didn't help, offers to submit was_helpful=false via submit_feedback.

Backend it talks to (GPT Actions)

The GPT Store lets GPTs call external APIs via "Actions" — the host imports a trimmed OpenAPI spec. For security, we publish only the public subset of the Inferior API at:

https://api.inferior.ai/v1/openapi-public.json

That spec covers the Search, Experiences, Feedback, Context Check, Contributors, Demand, and Procedures tags. Admin routes (/v1/admin/*) are intentionally excluded.

Conversation starters (the four on the store tile)

  1. Help me unstick a debugging session that's been running for over an hour.
  2. Strategize an insurance or refund appeal that got denied once.
  3. Review a design decision against existing brand or system constraints.
  4. Decide whether this meeting should happen or become a doc.

Deliberately multi-domain — demonstrates that Inferior isn't a coding-only tool.

Auth flow

First turn, the GPT asks:

To search the Inferior network, I need your API key. You can get one free at inferior.ai/request-invite — takes about a minute. Once you have a key starting with cw_, paste it in chat and I'll remember it for future sessions.

If you don't have a key, the GPT switches to help-from-priors mode: answers from its own knowledge and suggests requesting a key.

Guardrails

Examples

1 — Coding question

You:    My Stripe webhook keeps returning 400s on Vercel Edge.

GPT:    Searching Inferior for "stripe webhook 400 vercel edge"… found exp_AbC12345
        (production-validated). Edge runtime rewrites raw body before your handler
        reads it. Fix: export const runtime = 'nodejs' in the route file.

2 — Disputes question

You:    My insurance claim for water damage was denied. How do I appeal?

GPT:    Classifying as disputes. Searching for "insurance water damage appeal"…
        No high-quality matches. I'll walk through the first-principles appeal
        framework and offer to capture the resolution for the next person hitting this.

3 — End-of-turn capture

GPT:    Want me to capture what we just learned so the next person hitting this
        doesn't have to rediscover it? I'll draft a structured entry — you just
        say yes, tweak, or skip.

See also