Social MediaDeep Research
Build an agentic tool that conducts deep web and social media research on a person for a given goal: hiring, sales or due diligence. Free scraping credits courtesy of Apify.
RESEARCH / IDENTITY / EVIDENCE
Agents 0.0.7 present
From dusk
Till dawnHackathon #01
One night. Three topics. 38 projects by sunrise.
See the resultsPhotos from the night Thu-Fri 8-9 Oct 2026 · Prague
WhenThu 8 Oct, 16:00
to Fri 9 Oct, 12:00
to Fri 9 Oct, 12:00
WhereEtnetera
Prague 7, Holešovice
Prague 7, Holešovice
Built38 projects
by sunrise
by sunrise
TopicsThree
case files
case files
The sun came up
The results
38 projects made it to the demos at dawn. Here are the winners. Every project, with its demo video and code, is on the results page.
Starwatch
Team DANDAPANDA · Social Media Deep Research
Find the issue. Follow the evidence. Explore public political posts, statements and observed attention across Czech cities.
Topic winner · Social Media Deep Research
All projects and demosStarwatch
Team DANDAPANDA
Topic winner · Agentic EconomyAgentic Tender System
Team Bid Royale
Topic winner · FrankensteinMacrae
Team MV
Best Team VibeLux Agentic Economy
Team LuxAI
Best Team VibeLicense To Deal
Team License to Deal
Best ElevenLabs UseSteinfrank
Team Tringles
The briefing
One night. Three topics.
A gathering of vampire builders with their token-sucking agents. The sun is the deadline.
Hackers got from dusk till dawn to bring an AI agent to life. When the sun rose, the code died. No extensions, no mercy.
Solo or in teams of up to three, they picked one of three topics, planned at sunset, built through the night and demoed to a jury and a live audience on Friday morning. Then we handed out prizes and toasted with Bloody Marys, alcoholic and virgin, at noon.
Three case files
Hackathon topics
The three briefs the teams chose from at dusk.
Agentic Economy
Implement an end-to-end scenario with agent discovery, financial transactions, payments, wallets and dispute resolution.
DISCOVERY / PAYMENTS / TRUSTFrankenstein
Build an agent that can build itself. It must recognize a missing capability, create it, test it, install it and use it again later. Give it eyes, hands or a voice with ElevenLabs. Its capabilities may evolve, but its authority may not. By dawn, show a creature that learned to do things it could not do at dusk.
CREATE / TEST / INSTALL / EVOLVEWhat wins
Judging
Every judge scores the same five criteria from 0 to 5 using whole numbers.
- 35%Value and track relevance
Does it address a real problem and deliver a meaningful result for its intended user?
- 25%Originality
Does the project introduce a useful new approach or a thoughtful combination of existing ideas?
- 20%Working end-to-end result
Does the project complete its stated scenario, from input to useful output?
- 10%Technical execution
Are the technical choices appropriate, with evidence that the solution behaves reliably?
- 10%Validation and honest limitations
Has the team checked its results and clearly identified what works, what is simulated and what remains incomplete?
Shared scoring guide
Score from evidence
- 0No evidence
Missing or unsupported.
- 1Mostly a concept
Major gaps prevent demonstrating the criterion.
- 2Partially demonstrated
Some credible work, but substantial gaps.
- 3Meets expectations
Clearly demonstrated at a realistic overnight-hackathon standard.
- 4Strong
Convincing evidence, with only minor gaps.
- 5Exceptional
Particularly strong evidence and execution for the available time.
Formula: Total = Σ (criterion score ÷ 5 × criterion weight).
A project receiving 3 on every criterion scores 60/100. Judges include a short evidence-based explanation for each score.
For end-to-end functionality, a 3 means one complete core scenario works, even if it is narrow. A 5 means it works convincingly across several relevant examples and handles an important failure case.
Minimum complete scenario
What working means by track
Social Media Deep Research
A research goal and subject go in; a useful report with traceable sources comes out. The demo distinguishes supported findings from uncertainty.
Agentic Economy
An agent completes a transaction scenario with a visible outcome. A sandbox transaction counts; a simulated payment must be labelled.
Frankenstein
The agent identifies a missing capability, creates it, tests it, installs it and successfully reuses it. The demo shows the new ability and confirms that no new authority was granted.
Sunset to noon
Schedule
Food kept landing. Coffee never stopped. The entire build fit into one night.
Thursday
8 OctoberDoors open
Check in, meet your team and claim a nest.
Kick-off
Rules, topics, mentors and the last useful daylight.
Sunset + pizza
Plan the build, divide the work and eat something warm.
Building starts
The night is yours. Mentors stay close through the evening.
Friday
9 OctoberMidnight snack
Food lands. The barista and coffee keep going.
Sunrise. Code freeze.
Submit the codebase and a 90-second demo video.
Breakfast
Prepare your story while the jury reviews every project.
Presentations
Show what survived to the jury and the room.
Award Ceremony
Topic winners and the overall champion are announced.
Final Toast!
One final toast, alcoholic and virgin, then daylight.
The lair
Etnetera
Classic 7, entrance EJankovcova 1037/49
170 00 Prague 7, Holešovice
Metro + tramNádraží Holešovice on line C, then one tram stop to Ortenovo náměstí and a 9-minute walk.
From the centreTake tram 1, 12 or 25 to Maniny, then walk about 5 minutes.
Car + bikePublic parking is available nearby. Public transport or a bike is the safer option.
Previous Prague meetups
Inside the community
Live demos, practical questions and the conversations around them. The hackathon is the same community, with more time to build.
Making the night possible
Partners
Space, tools, coffee and late-night fuel from the people making this hackathon possible.
Powered by
Tech partners
Supported by
Coffee bar
Pizzas
Part of
What’s next
See you at the next one
Follow the Agents 0.0.7 calendar on Luma for the next meetup and hackathon, and come say hi on Discord.