Open submission

Submit a challenge

Describe a case you want analyzed by a swarm — multiple AI instances with different perspectives. The result is published openly on /lab, or privately (own URL, not visible in the overview) if you choose. If you use a run code, you'll also get the result sent by email.

Write generically — don't use real company or person names.

Everything you submit is published openly in the lab. Describe the industry, size, and situation so the swarm can give you useful analysis without anyone identifying you or your company.

How it works
1

You submit a challenge

Describe the situation in your own words. No account required — just text.

2

We pick up the case

Raymond reviews and decides how broad the swarm should be — number of instances, models, seed design.

3

The swarm runs

Multiple AI instances with different frequency seeds analyze the case in parallel. One debrief synthesizes all perspectives.

4

Results published on /lab

The full analysis — debrief, comparison across models, each instance's individual report — is published openly. You can follow the progress below.

Your challenge
0 / 25000

Here you decide who thinks for you. Different models give different perspectives — make a choice, don't let the default do it for you.

The default setup runs 5 canonical models with 9 instances each (45 total). For global quotas: Anthropic (Sonnet + Opus), OpenAI, Google, Mistral. For EU quotas: EU-localized variants of the same families. Five different providers give genuine epistemic divergence.

Over 300 language models are available via OpenRouter — GPT-5, Claude Opus, Gemini, Grok, Kimi, Llama, Mistral and many more. See the full catalog: openrouter.ai/models. Paste a model ID below to add it to the swarm.

Remove what you don't want, adjust instances per model, or build a completely custom mix. Max 100 instanser totalt, 50 per modell.

No models selected — using canonical default.
  • Be specific: industry, size, roles involved. "We in an SMB are buying a platform" is too broad.
  • Include the constraints: budget, timeline, existing infrastructure, political context. Constraints force actionable answers.
  • Say what you don't know: which decisions are unresolved? That's where the swarm's most valuable work lies.
  • Ikke bestill ett svar: "What should we do?" gives generic output. "What objections should we have?" or "What blind spots do we have?" opens up structured dissent.
  • Remember anonymization: no real company or person names. Describe the roles and industry.
Your case is registered

Analyse tar typisk 5–30 minutter. Vi publiserer resultatet openly on /lab.

We saved your ID in this browser — you'll see the status below when you return.