Open Civic Records
Public records are public in name only — buried in scanned PDFs and dead portals. An AI layer that makes them genuinely searchable, so anyone can find what's theirs.
Rebar Labs
Rebar Labs is where ambitious AI ideas get built — not theorized. Side Quests are real, funded projects you can join and build alongside others. The strongest don't stay side quests for long — they grow into main quests.
On client work, Rebar is ruthless about where AI belongs — most of it is better off deterministic. Labs is the other half of the brain: here we get to ask what happens if we push AI all the way, and aim it at something good.
Real budget, real backing from day one. No spec work, no equity-only asks — Side Quests get the resources to actually ship.
A think tank that ships. Side Quests get built, not whiteboarded — on Rebar's own platform and tooling so you focus on the idea, not the plumbing.
The best Side Quests don't stay side quests for long. They graduate to main quests — full-fledged products with the people who built them along for the journey.
Public records are public in name only — buried in scanned PDFs and dead portals. An AI layer that makes them genuinely searchable, so anyone can find what's theirs.
A swarm of cooperating agents that plan, divide, and execute multi-step operations end to end — pushing how far autonomous AI can actually be trusted.
An AI receptionist and scheduler for small service businesses — answering, booking, and triaging around the clock. A Side Quest with a clear path to a product.
A free, adaptive learning companion that meets each student exactly where they are — patient, unlimited, and built to push every learner as far as they can go.
Modern farming sprays first and asks questions later. Computer vision that watches the field plant-by-plant — spotting pests and weeds where they actually are, so the response goes only where it's needed and most of the field stays untouched.
Every business has its real know-how scattered across email threads, Slack DMs, dusty docs, and a few key people's heads. A queryable layer that consolidates it into working memory — so agents, and people, can act on what the company actually knows.
Most research time goes to running experiments by hand. A self-driving research loop — proposes hypotheses, designs experiments, reads the results, iterates — so scientists spend their time on the hard questions, not the protocol.
Most software gives you one UI for every kind of user. A coding agent that reshapes the interface around how you actually work — moves the buttons, drops the noise, surfaces what matters. Same tool underneath, refit for each person.
Conservation teams collect millions of hours of camera-trap footage and only ever look at a sliver of it. Computer vision that processes the whole archive — IDs species and individual animals, flags rare sightings — so the work that used to take a season takes a weekend.
Therapy happens in 50 minutes a week; the rest of the week is the hard part. A voice-first companion for the in-between hours — clinician-supervised, structured around real practices, designed as a bridge between sessions, not a replacement for them.
Every team shipping an AI agent has the same problem: no real adversarial coverage. A red-team that runs continuously against your live agents — prompt injection, tool misuse, data exfiltration — and ships a runtime policy bundle that closes whatever it finds.
Today's agents do impressive things in five minutes and fall apart in five days. A harness built for the long horizon — multi-session memory, scheduled wake-ups, self-evaluation, recovery from interruption — so an agent can carry a task across weeks the way a person would.
A panic attack doesn't wait for next week's session. A wearable-triggered companion that notices when your body says something's wrong — elevated heart rate, shallow breathing — and walks you through a grounding protocol in the moment. Not a replacement for care; a hand on the shoulder when you need one most.
Practicing alone only gets you so far. AI musicians that play along in real time — a drummer who follows your tempo, a bassist who locks in with you, a horn section when the song asks for one — so a bedroom musician can rehearse like there's a band in the garage.
Every gym mistake is the same — bad form you can't see from inside your own body. Phone on the floor, AI vision watches your squat or your swing, and a coach's voice in your ear corrects the angles in real time. The thing a good personal trainer does, finally available to anyone.
First-year parents wake up at 3am with questions a book can't answer and a pediatrician won't call back about until morning. A voice copilot that knows your baby — feeds, sleep, milestones, what the doctor said last visit — and is awake whenever you are.
Most people throw out a third of what they buy because they forgot it was in there. Snap a photo of the fridge — AI knows what you have, learns what you like, and lines up three dinners you could actually cook tonight.
The best books are better with someone to talk to about them. A reading companion that meets you where you are in the chapter — surfaces context, asks the questions a good professor would, lets you push back on the author — without ever spoiling what's ahead.
Most people have melodies stuck in their head with no way out. Hum into your phone — AI takes the seed and builds the song around it: chord progression underneath, beat behind, a full verse-and-chorus arrangement around the hook you didn't know you had.
If there's an idea where AI could genuinely make things better, pitch it. The strongest ones get backed, built, and grow.