visualizing the shift: pattern matching vs deterministic execution
speed0.5x
System 1 uses heuristic guessing.
given
solved
backtrack
propagate
error
System 1
pattern matching
steps 0
errors 0
restarts 0
confidence --
LLM Moves
System 2
deterministic compute
cells 0
consensus 0
backtracks 0
desperation 0
messages 0
Swarm Activity agent communication feed
waiting...
Agent Roster 16 autonomous agents
Row Scanners
Column Scanners
Box Scanners
Specialists
consensus 0%strategy idle
--16
Model
Model not loaded
Runs entirely in your browser. Downloaded once, cached locally.
How it works
Six AI agents with distinct perspectives — pessimist, optimist, analyst, challenger, integrator, forecaster — debate your question using a 1.5B parameter LLM running entirely in your browser. They extract entities from your seed material, build a knowledge graph, generate context-aware personas, argue over evidence, challenge each other's reasoning, and converge on a trust-weighted consensus prediction. Inspired by the MiroFish swarm intelligence approach.
Input
Knowledge Graph extracted entities and relationships
Progress
Initializing...0:00
tok/s: --phase: --total: --
Debate Feed
Agent Roster
Sentinel
pessimist
trust
claims: 0
idle
Catalyst
optimist
trust
claims: 0
idle
Historian
analyst
trust
claims: 0
idle
Contrarian
challenger
trust
claims: 0
idle
Synthesizer
integrator
trust
claims: 0
idle
Oracle
forecaster
trust
claims: 0
idle
Agent Interactions who challenged / supported whom