GTeam

The shared brain for your research lab.

Built at the YC GBrain × GStack Hackathon.

Every researcher has their own notes, papers, and findings scattered across their laptop. GTeam reads each researcher's personal GBrain, filters out the noise using Zero Entropy, and pulls only the lab-relevant work into one shared brain — so the whole team can search it, new hires can onboard instantly, and no one loses a week to something a teammate already knew.

---

What it does

1. Personal GBrains — each researcher has an AI (powered by GBrain) that captures everything they do: notes, papers, experiments, findings, hypotheses.

2. Selective pull — GTeam reads each researcher's personal files, scores them for lab relevance using Zero Entropy, and promotes only the work-related ones into a central shared database. Your grocery list stays private. Your fMRI findings go to the lab.

3. Paper agents — nightly agents scrape leading research labs and match new papers to each researcher's active projects using semantic embeddings + an LLM judge.

4. One searchable lab brain — any researcher can ask "what is Clara working on?" or "what papers relate to my motor BCI project?" and get an instant answer drawn from the whole team's knowledge.

---

Demo use cases

Use case	How
Onboarding	New hire asks the lab brain anything — project context, methods, who to talk to
Staying current	Daily digest of new papers matched and annotated per researcher
Team visibility	Dashboard showing what every lab member is actively building

---

Stack

Layer	Tech
Framework	Next.js 16 + TypeScript + Bun
Database	Supabase (Postgres + pgvector)
Embeddings	Voyage AI voyage-3-lite (1024 dim)
Relevance filtering	Zero Entropy (zerank-2 reranking)
LLM	Anthropic Claude (onboarding Q&A + paper–project matching)
Personal GBrain	GBrain via OpenClaw
Paper scraping	The Hog web scraper API
Email	Resend

---

Setup

cp .env.example .env.local   # fill in keys (see below)
bun install

# Apply schema to Supabase:
# Dashboard → SQL editor → paste db/schema.sql → run
# Then paste db/seed.sql → run

bun dev

Required environment variables

SUPABASE_URL
SUPABASE_ANON_KEY
SUPABASE_SERVICE_ROLE_KEY
ANTHROPIC_API_KEY
VOYAGE_API_KEY
ZERO_ENTROPY_API_KEY
HOG_ACCESS_KEY
HOG_SECRET_KEY
RESEND_API_KEY          # optional — falls back to console

---

Scripts

bun run brain:daily     # 5:30 PM Pacific Central GBrain source refresh + embeddings
bun run fetch-arxiv     # scrape leading lab publication pages → upsert papers
bun run match-papers    # match papers to researcher projects via LLM judge
bun run render-digest   # render per-researcher paper digest as markdown
bun run send-digest     # email digest to all researchers
bun run seed-corpus     # seed demo-data artifacts into Supabase

---

Project layout

app/          Next.js routes (/team, /onboard, /upload)
components/   React components
lib/          Supabase, embeddings, Anthropic, artifacts, email
db/           schema.sql, seed.sql, typed DB client
prompts/      onboarding Q&A and relationship judge prompts
scripts/      paper fetch, matching, digest, seeding
cli/          researcher GBrain sharing helpers
demo-data/    mock researcher artifacts for demo seeding
docs/         feature specs and migration notes

---

Team

Built at the YC GBrain × GStack Hackathon by a 4-person team in 24 hours.

• P1 — Infra, DB schema, Supabase, paper scraping
• P2 — Web app (Next.js routes, React components)
• P3 — AI pipeline (embeddings, LLM judge, paper matching)
• P4 — CLI, artifact ingestion, digest delivery