72 hours to build one thing and show that it works. Everybody competes from home. We read every application by hand and we turn most of them down.
A demo only its author can drive isn't a solution. That is the whole premise. You get 3 days, a single problem you actually care about, and an audience on Sunday afternoon that will ask what your thing does when the input is garbage.
All of it runs online. No venue, no flights. You work at your own desk on your own hours, with a team you bring or a team we put you in on Friday morning.
The group stays small on purpose. Judging 40 demos carefully takes longer than judging 400 badly, and we would rather do the first one.
You get mentors who have run this kind of system in production. You get judges who ask the question you were quietly hoping nobody would ask. And you get 3 days without meetings, which for most of us is the rarer prize.
Build something that survives contact with a user.
Pick one when you apply. Switch on Friday if your team drags you somewhere better.
Software that takes several steps on somebody's behalf and knows when to stop. What we care about is what yours does on the step that fails.
The part a person touches. A chat box bolted to the corner of an existing product is the low bar. Clear it.
Getting the right context in front of the model. Search, ranking, memory, structured sources, whatever it takes to keep the answer honest.
How you know it works. Then how you find out it stopped working, before a user has to tell you.
The layer everything else sits on. Extra credit for a pipeline that stays upright when somebody changes the input format at 2am and mentions it to nobody.
Anchor sessions get recorded so nobody loses a day to a timezone. Between them you build.
Briefs and teams
Build and break
Ship and show
This one isn't open to everybody. We don't count years and we have no opinion about your title. We do want to see a thing you built that somebody other than you has used.
A side project with 12 real users counts. So does a research prototype that somebody else has run without you sitting next to them. A repository with a beautiful README and nothing behind it does not.
Send the work first. We'll read that before we read anything else about you.
The first AI Solution Makers Hackathon ran on October 9 and 10, 2025. 2 days, fully remote, put together fast, mostly to find out whether the format held up at all.
It held. People turned up with a problem on Thursday morning and left on Friday evening with something running, and the demos that afternoon were sharper than what we've sat through at conferences 10 times the size. One complaint came back from almost everybody: 2 days was not enough.
So the second edition, on September 28 to 30, 2026, ran to 3 days and brought in a judging panel recruited months ahead rather than the week before. The extra day went to building. Nobody asked for more programming.
November keeps all of it. Small group, remote, no filler, a real deadline, and the same rule about who gets in.
Judges are the reason the demos are worth watching. We want people who have built this kind of system themselves and can tell a good demo from a good system inside of 90 seconds.
The commitment comes to a few hours. 1 briefing call in advance, a midpoint check-in with a couple of teams, then demos on Sunday. All remote, and you pick the tracks you're closest to.
Mentors do the less visible half of it. You hold office hours, somebody shares a screen with a stack trace on it, and you save that team half a day. 2 hours is a useful shift. Take the slots that fit around your actual week, on the subjects you know cold.
We are selective about mentors for the same reason we are selective about builders. A team gets a couple of hours of your attention and then goes and acts on what you told them, so bad advice costs them a day they don't have. We would rather run a short bench of people who have done the work than a long list of names.
If you've debugged this at 1am, somebody here wants to talk to you.
Each of the 3 days opens with 1 short talk. 20 minutes, live, on something you have built or broken yourself. Anybody can put their name in.
There is no stage and no keynote energy. What you get is a room of people who are mid-build and will come straight at you with questions when you finish. The talks that land are specific: a system you shipped and what it cost, a decision you would take back, a failure mode nobody warned you about.
We pick these the same way we pick judges and mentors, which is to say carefully. A talk takes 20 minutes from every person at the event, so we would rather run 3 good ones than fill a schedule.
Pitch usConfirmed so far. We announce the rest as they lock in.
One short talk opens each of the 3 days. We announce the line-up as speakers confirm.
The same address, whichever way you want in. Write to us. Every email gets read and every email gets an answer, including the nos.
Bring a team or come alone and we'll place you on Friday morning.
A few hours across the 3 days, remote, on the tracks you know best.
Office hours in whatever slots suit you. The bench is short and we read these applications the same way we read the builders'.
20 minutes on the morning of your choice. 3 slots across the event.
Either way, the address is hello@aisolutionmakers.com
Yes. Every session happens online and there is no in-person component at any point. You need a laptop and a connection you trust.
Plenty of people do. We run team formation on the first morning. You can also compete alone if that suits you better.
No, and please don't. Anchor sessions are recorded and mentor hours are spread across timezones. Work the hours your body actually keeps.
You do. Your team keeps the code and everything around it. The only thing we ask is that you demo it publicly on the last day.
Turn up with a stack you know and any boilerplate you'd reach for on a normal project. The solution itself gets built during the 3 days.
20 minutes, once a day, live at the start of the morning. Slides are optional and a screen share of the actual thing is usually better.
No. Mentor applications get the same read as builder applications, and we turn plenty of them down. Teams act on what a mentor tells them, so we take on fewer people and pick ones who have shipped the thing they're advising on.
Yes. Say so in your email and tell us which one you'd rather do if we have to pick.
We review as they arrive and stop when the group is full, so earlier is better. Email us and we'll tell you honestly where things stand.
Bring back something that works. November 20-22, 2026, virtual, applications read as they arrive.
The people who open each morning with a talk, and the people who watch your demo on the last afternoon and ask what happens when it breaks. We recruit them months ahead and brief them before the event starts.
We are recruiting judges and speakers for the November edition now. Names go up here as they confirm, and every judge sits through a briefing call before the event so the teams get consistent scoring.
A briefing call, a midpoint check-in with a couple of teams, then demos on Sunday. All remote.