Swarmie
Founder validation through grounded AI swarm simulation. Paste a pitch, watch 500 AI personas react like a Reddit thread, and walk away with the three real objections before you burn a single user interview.
- 500
- Max agents
- ~60s
- Pipeline runtime
- $0
- Local Ollama cost
The problem and the build
Founders spend $10k on user interviews, or six months on the wrong positioning, to discover the same three objections a roomful of strangers would have raised in fifteen minutes. Swarmie inverts the loop. Paste a pitch and the system parses ICP, problem, solution, pricing. A swarm of 100 to 500 LLM driven personas reacts in character. Post, comment, upvote, ignore. Each with distinct tones and objection biases. A reporter synthesizes top objections, sentiment split, PMF score, per ICP fit, and messaging gaps. Click any reacting agent, see its persona and reasoning, then chat with it to dig deeper. The brain view is a live d3 force canvas where 240+ neurons orbit archetype satellites, pulsing amber when their LLM call fires and settling to green or red based on sentiment. Two tier model routing keeps costs under control. 60% of agents ignore and cost zero tokens, matching real social media base rates. A watchdog cancels mid run if spend exceeds a hard ceiling. The whole loop lands in about 60 seconds.
Key insight
Selection before simulation. Most validation tools summarize or poll. The leverage is in letting agents perform the behavior, not predict it. Action roll before LLM call, tiered routing, and a hard cost cap turn a swarm into something you can actually run on free tier.
Stack
- Vue 3
- Flask
- OpenAI-compat LLMs
- d3-force
- Remotion
More case studies
- LearnAIA free map of all of AI, from linear algebra to agents. 14 fields, 152 steps, 162 problems. Every step is something you read, then something you build. Finish it and a star lights on your map.
- PitchcraftA pipeline that turns a CSV of local businesses into deployed prototype websites, audit PDFs, and growth proposals, ready to send. One operator, a batch of leads, a folder of finished deliverables.
- Bubblesheet AIGrades multiple choice exams from ordinary scans. A YOLOv8 model reads every filled bubble, checks it against the answer key, and returns per student scores as a CSV.