DopaPal
A cognitive translation layer for the ADHD brain.
Ranking
PINCH
Surface
Desktop (Electron)
Brain
Nemotron LLM
Metric profile
01 / Signal
<5s
Capture friction
voice, hotkey, clipboard
02 / Signal
5 inputs
PINCH signals
passion → hurry
03 / Signal
Micro-blocks
Task slicing
paced, dopamine-aligned
Overview
Traditional task managers assume a brain that responds to deadlines. DopaPal assumes one that responds to interest, novelty and momentum — and builds the entire scheduling model around that.
Capture is deliberately frictionless: voice, global hotkeys, clipboard and calendar sync. Overwhelming projects are auto-sliced into paced micro-blocks small enough to start without negotiation.
What makes it work
PINCH ranking
Passion, Interest, Novelty, Challenge and Hurry combine into a score that surfaces what you can actually start — urgency is one input, not the whole model.
Auto-slicing
The LLM decomposes vague, oversized projects into paced micro-blocks with explicit first actions, removing the activation-energy wall.
Dopamine-aligned rewards
Streaks, momentum feedback and variable reward pacing are tuned to reinforce follow-through rather than punish drift.
Architecture
- Electron + React desktop shell with global hotkey capture
- FastAPI backend orchestrating slicing and ranking
- PostgreSQL for task state, Redis for real-time queues
- Nemotron LLM for decomposition and prioritisation
Stack
- Electron
- React
- FastAPI
- PostgreSQL
- Redis
- Nemotron LLM
An ambient cognitive translation layer for the ADHD brain. DopaPal captures chaos through low-friction input (voice, hotkeys, clipboard, calendar sync), auto-slices overwhelming projects into paced micro-blocks, and surfaces work by PINCH ranking (Passion, Interest, Novelty, Challenge, Hurry) instead of pure urgency — wrapped in a dopamine-aligned reward engine.
Interested?