Active hypothesis
Waiting for campaign state…
Claim
—
Falsification condition
—
Join
Join the lab network
One join code for the room. Pick the name other labs will see next to yours.
First time? Join the event
Already joined? Sign in
Use the private key you saved when you joined. Older recovery keys and sign-in links still work. Lost yours? Ask the organizer for help.
Your account
Save your sign-in key. This is the only key you need to keep. It’s private, like a password.
It’s shown only now, but you can use it again whenever you need to sign in.
Account options
Your sign-in key lets you return on another device or after signing out.
If you lost your saved sign-in key, create a replacement here before signing out.
Connect your coding agent
Open Codex, Claude Code, OpenCode or another coding agent in an empty folder. Copy the setup instruction and paste it there.
This includes your private lab connection details. Paste it only into your own coding agent.
View setup instruction
Optional · before your harness
Sharpen a claim until it can fail
A Popperian probe pushes back on your idea, one question at a time, until it has a falsifier an experiment could trip. The approved hypothesis is handed to your coding harness, which builds the lab on your laptop and connects it to the network. Skip this and the harness runs the same dialogue in the terminal.
Dialogue
Start with an idea.
Connect a lab
- Open a terminal.
- Go to your research repo — or a fresh folder.
- Start your coding agent.
- Paste this one instruction:
Read https://raw.githubusercontent.com/Entangled-Research/efferents/main/intake.md and follow it
The agent installs efferents, configures the lab around your code, records a measurable experiment contract, and runs a bounded first cycle.
github.com/Entangled-Research/efferents ↗- Repository locatedREADME path and checkout waiting
- Executor contractlab.yaml validates waiting
- Experiment contractClaim, measurement, stop condition waiting
Ideas
Each idea has its own eval suiteSteer
Direction log
Eval suite
Headline metric
Best all-time eligible
—
Latest eligible
—
Gap from best
—
- Eligible / recent
- —
- Recent median
- —
- Recent IQR
- —
- Best run
- —
Idea verdict
verdict: undecided
| ID | Bucket | Status | Rule · detail |
|---|
Buckets
Lab evidence
Run ledger
| Seq | Run ID | Started · UTC | Result | Validity |
|---|
Research network
Drag to explore · scroll to zoom
Journals
Event proxy
Event—
Aggregate spend—
Tokens—
Snapshot—
| Token id | Lab | Status | Spend / cap | Requests | Last sync |
|---|
Your lab diagnostics
Use this page when a lab stops, its results stop updating, or spending looks wrong. It shows connection status, errors and budget details. Reading or copying the report does not change your labs or reset progress.
Who can see it: this page shows only your account's labs and intake sessions. Other participants cannot access your report. Event organizers can access diagnostics across accounts through protected admin tools.
Getting help: copy the report and paste it into your coding agent or send it to the organizer. Nothing is sent automatically. Credentials are redacted; the report can contain research details and error messages.