Free beta
Discover ideas.
Decide with evidence.
Cogitate is two jobs, not a dozen settings. Discover invents non-obvious ideas and tells you which are cheap to test. Decide helps you choose among options you already have.
Bring at least one LLM provider key (OpenAI, Anthropic, Google, or xAI — more is better). Typical runs are about $0.50–3 on your provider bill.
You describe what you want to invent. Models invent in parallel — as many providers as you have keys for — then Cogitate sorts the results into worth trying, needs homework, park, and not now, with a short test plan for the keepers. Creativity first; proof later.
You already have options. Paste the facts and constraints. Cogitate builds an evidence-bound recommendation with kill criteria — not a vibes-based answer from a single chat.
How a run works
-
You provide the brief
Type or dictate. Add notes for constraints, numbers, and prior decisions. Voice notes can be auto-tidied before the models read them.
-
Pick Discover or Decide
One choice. We handle the models, rivalry framing, and review stages behind the scenes.
-
Read the result
Discover returns a shortlist of ideas worth trying. Decide returns a recommendation you can act on — or challenge with a follow-up.
Why not one chat window?
A single model gives you one attractive answer. Discover forces several inventors to compete, then ranks what is actually testable. Decide forces evidence and kill criteria before a recommendation is allowed to stick.
What you need before signing up
Bring your own provider keys. Cogitate orchestrates; it does not pay for tokens. You add your own OpenAI, Anthropic, Google, or xAI keys and those providers bill you directly. One key is enough to run; more keys improve the diversity of the drafts.
It is not zero-knowledge. Stored keys are encrypted at rest, but the server decrypts them in memory at run time to call the model provider. Use fresh, revocable, budget-limited keys.