ExponenLabsExponenLabs
Founding Pilot · Free · No equity

Building a startup alone is hard. Don’t make the product and tech calls alone.

For six weeks, senior builders work next to a few early-stage founders on both halves of the job: what to build and how to build it, the AI-native way. Free. No equity. You keep everything.

See the six weeks

Five questions · about 3 minutes · we reply within 7 days, yes or not yet

Starting with 2–4 startupsZero equity takenYou own 100% of the IP & code

Sound familiar?

You have a real problem you care about, but you are not sure what to build first.

Your prototype works on your laptop and breaks the moment a real user touches it.

You are making product and tech decisions alone, with nobody senior to check them.

Every advisor wants equity or a retainer before they will give you a straight answer.

This pilot is for exactly that moment.

Why is this free? The honest answer.

Most accelerators take 7–10% of your equity before you have proved anything. We do not want your equity, and we do not want to bill founders who are bootstrapping.

This is our first pilot. We are building a practice around helping early-stage founders, and working closely with a few of them sharpens our tools and our own thinking. You will shape the program with us, and we will learn as much from you as you do from us. If your company later wants ongoing senior help, that is a separate conversation, and the answer can be no.

Whatever you are building, we meet you where you are

Most founders we expect are building AI products. But AI changes how every company gets built and run, so you do not need AI in your product to apply.

Early idea

From a real problem to a clear plan

Who it is for: You have spotted a painful problem and talked to a few people about it, but you do not yet know what version one should be.

What we do: We work out who the first users are, what version one will not do, and what to build first, with a working prototype you can show them.

Rough prototype

From a fragile demo to a working MVP

Who it is for: You have a script, notebook or demo that sometimes works, or costs too much, or breaks on edge cases.

What we do: We make it reliable and affordable, then put it in front of your first real users and learn from what they do.

Not an AI product

From a normal startup to an AI-native one

Who it is for: You are building SaaS, a marketplace, or something with hardware or robotics, where AI is a critical component or not in the product at all.

What we do: We show you how to use AI to research, build, support and operate the company faster, and where it belongs in the product.

The 6-Week Sprint

How we work together, week by week

Every phase has a product track and a tech track, because a well-built product nobody wants is as bad as the reverse. AI-native habits are woven through both.

Weeks 1–2
Phase 1

Make sure you are building the right thing

Product track

  • Talk to 5 people who match your target user
  • Write a one-page problem statement
  • Decide what version one will not do

Tech track

  • Decide what AI should do and what plain software should do
  • Choose models and tools you can afford to keep
  • Sketch the architecture and the data your product needs
Technical detail

We split the work between fast classification (System 1) and deeper reasoning (System 2), and choose between models such as OpenAI, Claude, Gemini or small typed models like Jev. You leave with a technical architecture brief and an API data contract.

Weeks 3–4
Phase 2

Get something real working

Product track

  • Pick the one workflow a user must be able to finish
  • Check in with those users every week
  • Keep a short, ranked list of what to build next

Tech track

  • Build a working end-to-end prototype
  • Set up an AI-assisted way of building that your team can keep using
  • Test it against realistic examples so you can measure quality
Technical detail

The prototype includes prompt templates, retrieval over your own data (RAG), a tool-calling harness with state machines for agent steps, and a synthetic test fixture to measure accuracy against baseline queries.

Weeks 5–6
Phase 3

Make it reliable and put it in front of users

Product track

  • Run your first 5–10 real user sessions
  • Read the signals and decide what to do in the next 90 days
  • Know what to say to early customers and design partners

Tech track

  • Set spend limits and fallbacks so nothing runs away or fails silently
  • Add automatic checks before results reach users
  • Hand over a clean repository and a prioritized technical roadmap
Technical detail

Guardrails include hard spend caps, rate limits, timeout budgets and a kill switch. Evaluations and automated verification run before results reach users, and telemetry records failure traces from the first real runs.

Open-source tools you can use if they fit

You do not start from a blank canvas. Founders in the pilot get plug-and-play access to the same tools we publish in the open.

Jev Playground

Quick, cheap decisions in under 200ms, so simple tasks do not burn expensive AI model calls.

View tool details

AI Engineering Cookbook

Proven prompts, agent skills and review guardrails so AI-assisted coding stays reliable.

View cookbook

Hybrid Search RAG

Lets AI answer from your own data instead of guessing, by combining meaning-based and exact-keyword search.

View RAG tool

Radical focus: what we do and what we skip

Early on, doing too much is the most common way to fail. We stay focused on proving that something people want works.

What we do
  • Work through product decisions with you: who it is for, what to build first, what to cut
  • Work through technical decisions with you: architecture, AI versus plain software, tools, cost
  • Show you how to run the company AI-native, whether or not AI is in your product
  • Review and co-design your code, and unblock technical dead ends
  • Help you put your first test users on the product and learn from them
What we deliberately skip
  • Premature scalability, multi-region clusters or microservices overkill
  • Polished mockups with nothing real working underneath
  • Enterprise compliance work before you have found a problem worth solving
  • Taking any equity, cap table stakes or intellectual property

What we ask of you

A real problem: a concrete situation that genuinely hurts someone.
About 6–8 hours a week for six weeks, including a weekly working session with us.
Talk to real users by week three, even if it feels early.
Honesty: tell us what is not working. If it is not working by week two, we will say so and part as friends.

Frequently asked questions about the bootcamp

Everything you need to know about the pilot, who it is for, and how to apply.

Is this really free? What is the catch?
There is no catch. We do not charge a fee, take equity, or take any intellectual property. This is our first pilot. We are building a practice around helping early-stage founders, and working closely with a few of them sharpens our tools and our own thinking. If your company later wants ongoing senior technical help, that is a separate conversation and the answer can be no. If not, you leave with what you built and owe us nothing.
I am not technical. Can I apply?
Yes. Domain experts with a real problem and a builder mindset are welcome. Half of the program is about product decisions: who the user is, what to build first and how to learn from early users. We handle the technical side with you and explain every decision in plain language.
Do I have to be building an AI product?
No. If AI is the core of your product, we help you get it right. If AI is a critical component, as in robotics or hardware, we help you decide where it belongs. If AI is not in your product at all, we show you how to build and run the company AI-native: using AI to research, build, support and operate faster.
What stage do I need to be at?
Anywhere from a clear problem and some customer conversations to a rough prototype or an MVP with a few users. It is not for scaled companies optimizing existing traffic.
How much time will it take?
Plan for about 6–8 hours a week for six weeks: a weekly working session with us plus your own build and user time. This is a builder sprint, not a lecture series. If it is not working by week two, we will say so and part as friends.
How many startups will you take?
This is a pilot, so we are starting small, with 2 to 4 startups. That way every team gets direct senior attention. Applications are open on a rolling basis, and we start when we have found the right teams rather than on a fixed date.
Who owns the code and the IP?
You own 100% of your code, models, data and intellectual property from day one. Everything runs in your repository and accounts.
How does applying work?
Fill in the short form: five questions, about three minutes. We read every application and reply within 7 days, whether it is a yes, not yet, or a pointer to something useful. If it looks like a fit, we will set up a short call.
What happens after the six weeks?
You keep the working product, the repository and the roadmap, and you are free to carry on alone. If you want continued help, we can talk about it then. There is no obligation.
Applications open · reviewed on a rolling basis

Got an idea and no one senior to think it through with?

Tell us what you are building and what is in your way. Five short questions. We read every application and reply within 7 days, whether it is a yes, not yet, or a pointer to something useful.

About 3 minutes · No obligation · No equity, no fee, you keep all IP