Daimon, the Bayesian data science agent from PyMC Labs

Daimon Hackathon · PyMC Labs · Date and venue to be announced

Three tracks. One agent. Ship it.

Analyse a company's data, publish a data story, or land a pull request in the PyMC libraries — with Daimon as your teammate. Every participant gets $25 in Daimon credit. One judge picks what holds up.

  • 3
    Tracks
  • 4–6
    Per Team
  • $25
    Daimon Credit
  • 1
    Judge

Presented by

  • PyMC Labs
Chapter I

The Brief

The three tracks and who does what on a team of four to six.

  • Objective
  • Tracks
  • Roles
The Challenge

Pick a track. Put Daimon to work. Ship something a judge can open.

Three tracks, one agent. Analyse a company's data warehouse, publish a data story from public sources, or land a pull request in the PyMC libraries — with Daimon doing the heavy lifting. Every participant gets $25 in Daimon credit. What you ship is what gets judged.

The Tracks

Pick a track, then ideate freely within it.

Three ways in. Pick the one that fits your team, then define the problem, the output, and what done looks like in your brief.

Track 01

Data Analysis

A fictional company's data warehouse: realistic tables plus call recordings. Use Daimon and the PyMC libraries to dig through it, build predictive models, and show what the business should do next.

For teams who want to model, forecast, and visualize.

Track 02

Data Journalism

Gather public data — SEC filings, government datasets, open APIs — compile and analyse it with Daimon, and publish a data blog post that links back to PyMC Labs. Stick to sources you can download; scraping gets blocked.

For teams who want to find the story in the numbers and tell it.

Track 03

Benchmarks & Features

Dig into the PyMC libraries and their code and data-analysis benchmark. Find a feature to add or an evaluation to improve, build it with Daimon, and open a pull request. Merged and well pitched is a win.

For teams who want to ship upstream and argue about design first.

One judge, one question: does the result hold up? A model that predicts, a post worth reading, a PR that gets merged.

Roles

Everyone has a job. Non-technical people aren't support.

Driver

Runs Daimon, owns the repo or the notebook, and keeps the thread moving.

Analyst

Owns the questions. Decides what to ask, checks the numbers, and calls out what does not add up.

Writer

Turns the findings into the post, the PR description, or the pitch the judge reads first.

Reviewer

Reads everything cold before it goes out. If it does not make sense to them, it is not finished.

Teams are four to six people, formed around a track, not a job title. One person can drive Daimon while the rest argue about what to ask it. That argument is where the good entries come from.

Chapter II

The Build

Six steps, one tool, and the walkthroughs. Everything between picking a track and Daimon running.

  • Steps
  • Stack
  • How-to
The Six Steps

A clear pipeline from track to submission.

  1. 00Pick

    Choose a track and form a team.

    Join the hackathon channel on the Decision AI Discord, find your people, and say which track you are in.

  2. 01Set Up

    Add Daimon to your server.

    One click in Discord or Slack. Hackathon participants get $25 in credit. Running the open-source version on your own keys is fine too.

  3. 02Frame

    Write down the question.

    What are you trying to find out, publish, or fix? One paragraph before anything runs. It becomes your pitch.

  4. 03Build

    Let Daimon do the work. Check it.

    Ask it to explore, model, chart, or code. Read what comes back. Push on anything that looks too clean.

  5. 04Finish

    Ship the artifact for your track.

    A notebook with charts, a published post, or an open pull request. Not a plan for one.

  6. 05Pitch

    Tell the judge what you found.

    A short pitch, read before the work is opened. Say what you set out to do and what happened.

The thread is the record — Daimon works in one Discord or Slack conversation, so the whole team sees every step.

The Stack

Tools that power every team.

Daimon lives in Discord and Slack. Scan the code or open the link to add it to your team's server.

DAIMON

Your fifth teammate

Data science agent from PyMC Labs. Writes and runs code, fits Bayesian models with PyMC, posts charts and runnable notebooks in the thread. Free to add. Hackathon participants get $25 in credit.

PYMC OPEN-SOURCE STACK

Optional · Special Prize

Before you start building, ask Decision Hub what agent skills already exist — it may save you hours.

Chapter III

The Day

The timetable, the rules, and where to register.

  • Schedule
  • Rules
  • Register
Schedule

Timetable coming soon.

Date, venue, and the running order are being confirmed. Join the Discord to hear first.

Rules

Eight rules. No exceptions.

  1. R01

    Daimon does the work. Use the managed version or run your own — either way the analysis and the code go through it.

  2. R02

    Teams of 4–6. No solo submissions. One track per team, named in your submission.

  3. R03

    Track 01 ships an analysis of the provided warehouse: findings, models, and charts.

  4. R04

    Track 02 ships a published data post that links back to PyMC Labs. Public sources only — downloads, filings, open APIs. No scraping.

  5. R05

    Track 03 ships a pull request to a PyMC library. Merged and pitched is a win.

  6. R06

    Show your work. Every finding comes with the code and data that produced it.

  7. R07

    One Devpost entry per team: track, repo or notebook, the post or PR link, and a written description.

  8. R08

    Submit on Devpost by the deadline — no extensions.

Register

Three steps to lock your spot.

  1. 01

    RSVP

    Reserve your seat on Partiful.

  2. 02

    REGISTER YOUR TEAM

    Two registrations — Devpost for your submission, the form so prizes reach the right team.

    QR code for https://vibeforward-m2-agents.devpost.com/
    Team verification form: Verify team ↗
  3. 03

    JOIN THE DISCORD

    Team formation, questions, and announcements happen in the hackathon channel on the Decision AI Discord. Add Daimon to your server while you are there.

Chapter IV

The Prize

How the judging works, what teams take home, and everything else you asked.

  • Judging
  • Prizes
  • FAQ
Judging

Five criteria. 25 points maximum.

  1. 01

    It works

    Ran on real data. Outputs aren't hardcoded.

    5 PTS
  2. 02

    Daimon did the work

    The analysis or the code came out of the Daimon thread, not just the write-up.

    5 PTS
  3. 03

    Matches your question

    Judged against the question you wrote down in Step 02.

    5 PTS
  4. 04

    Someone can use it

    A reader can follow the post, run the notebook, or review the PR cold.

    5 PTS
  5. 05

    Explainable

    Every finding has a visible reason a human can follow.

    5 PTS

Minimum 15 points to qualify for finalist consideration. One judge scores every submission. Finalists present their work live.

Prizes

Prizes to be announced.

The prize pool is being finalized and will be posted here once confirmed.

Everyone gets: $25 in Daimon credit, five times the standard install credit.

Track 03 has its own win condition: a pull request that gets merged into a PyMC library.

FAQ

Common questions.

Eligibility

I'm not technical — can I compete?

Yes. Track 02 is built for it: find public data, ask Daimon the questions, write the post. Track 01 rewards people who know the business better than the code. Track 03 is the one for engineers.

Tooling

Do I need my own API key?

No. Every participant gets $25 in Daimon credit, up from the usual $5. Prefer your own setup? The open-source Daimon runs on your own keys.

Data

Can I scrape websites for Track 02?

Don't. Scrapers get IP-blocked mid-hackathon and take your afternoon with them. Use downloads, SEC filings, government datasets, and open APIs.

Submission

What must I submit?

Your Devpost entry must include: track selection, a GitHub repo or notebook link, the artifact for your track (charts, the published post, or the pull request), and a written description of what you set out to do and what you found. All items required — partial submissions are not accepted.

Judging

How is the winner picked?

Five criteria, five points each, 25 maximum. Minimum 15 to be considered. One judge scores every submission and the finalists present live.

Still have a question?

Ask in the hackathon channel on the Decision AI Discord. Registration is open on both Devpost and Partiful.