Resources · 2026–27

What to read before you build.

Everything below is here because it changes how you approach the seven weeks. Founders worth studying, a week-by-week playbook, and the structures judges expect in your submission.

Founders worth studying

People who started where you are.

Three founders whose companies came from an actual problem rather than a clever idea. Every link goes to something they said publicly — talks, interviews, transcripts. Watch one properly rather than skimming all nine.

Alexandr Wang

Scale AI

Started at 19. Dropped out of MIT to do it.

Wang noticed that companies wanted to use machine learning but didn't have the data infrastructure that big tech had in-house. He didn't build a better algorithm — he built the boring, unglamorous layer underneath it that everyone else needed. Scale went from three employees to a company serving OpenAI, Airbnb, and the US government.

What to take from it

He picked a problem that wasn't cool yet. His advice to founders is to enter a market early, before it's obvious, and to start narrow before chasing the big market.

Hovhannes Avoyan

Picsart

Built it because his 10-year-old daughter was bullied.

Avoyan's daughter came to him in tears after classmates mocked her artwork online. He wanted her to have a place to make art without that. That became Picsart — now a creative platform with over 150 million monthly users. It was his sixth company.

What to take from it

The best problem to work on is often one you've actually watched someone you care about struggle with. He wasn't the obvious person to build a Gen Z creative app — he built it anyway because he understood the problem personally.

Daniela Amodei

Anthropic

Literature major. Congressional staffer. Now runs an AI lab.

Amodei studied literature at UC Santa Cruz and worked as a Capitol Hill staffer before moving into tech at Stripe, then OpenAI. In 2020 she co-founded Anthropic, where she is president. She has said the company deliberately hired across disciplines — political scientists, ethicists, biologists.

What to take from it

You do not need a computer science degree to build an AI company. Amodei has argued that human skills become more important in the age of AI, not less — and that judgment about what to build matters more than the ability to write the code.

These founders are not affiliated with IASC and have not endorsed the competition. They are listed here because their public interviews are genuinely useful to student founders.

Build-phase playbook

Seven weeks, week by week.

These are the optional milestones from the competition timeline, with what to actually do in each one. The build phase runs November 2, 2026 to December 20, 2026.

  1. 1

    Weeks 1–2

    Customer discovery

    Prove someone other than you has this problem.

    • Write down who exactly has this problem — an age, a job, a situation
    • Talk to at least 10 of them. Ask what they do now, not whether they'd use your idea
    • Record what surprised you
    • Rewrite your problem statement afterwards. It will change

    Common mistake — Asking 'would you use this?' Everyone says yes to be nice. Ask what they did last time they had the problem instead.

  2. 2

    Week 2

    Market validation

    Work out whether this could be a business, not just a project.

    • Estimate how many people have this problem, and show your working
    • List who already solves it — including 'they just put up with it'
    • Decide what you do differently, in one sentence
    • Sketch how money would eventually change hands

    Common mistake — Claiming a huge market with no source. Judges score the reasoning, not the number.

  3. 3

    Weeks 3–4

    MVP planning

    Build the smallest thing that proves the idea works.

    • Write the one thing your product must do. Cut everything else
    • Pick your tools — no-code and AI tools are fine and encouraged
    • Set a hard stop date. Ship something rough rather than nothing polished
    • Keep notes on what AI did versus what you decided

    Common mistake — Building for four weeks and showing nobody. Get it in front of a real user in week three.

  4. 4

    Week 5

    Product feedback

    Watch someone use it without you helping.

    • Sit five people in front of it and stay silent
    • Write down every point where they hesitate
    • Fix the top three problems only
    • Capture one quote from a real user for your pitch

    Common mistake — Explaining it while they use it. If it needs explaining, that's the finding.

  5. 5

    Weeks 6–7

    Pitch practice

    Be able to explain the whole thing in three minutes.

    • Write the script before you record. Time it
    • Record once, watch it back, then delete it and record again
    • Practise the founder Q&A unscripted — no notes
    • Have someone who knows nothing about it watch and tell you what they heard

    Common mistake — Spending week seven on video effects. Judges score clarity, not production.

Submission guide

Three minutes, structured.

Not a rule — a structure that works. Every second you spend on something outside this is a second not spent on what's scored.

Suggested pitch video structure by timestamp
TimeSection
0:00–0:20The problemWho has it, and why it matters. Be specific about the person.
0:20–0:50Your solutionWhat you built, in plain language. No jargon.
0:50–1:20Why AIWhat AI does here that nothing else could. This is 20% of your score.
1:20–1:50Who it's forYour actual target user, and how you know they exist.
1:50–2:20The businessHow it makes money and how big it could get.
2:20–2:45Your evidenceA demo clip, a user quote, a number — anything real.
2:45–3:00The visionWhere this goes in five years, and who's on the team.

Pitch deck — 8 to 10 slides

One idea per slide.

Optional, but it helps judges follow your reasoning. If a slide needs a paragraph to explain, split it.

  1. 01Problem
  2. 02Solution
  3. 03AI technology
  4. 04Market opportunity
  5. 05Product
  6. 06Business model
  7. 07Competitive advantage
  8. 08Go-to-market
  9. 09Vision
  10. 10Team

Founder Q&A — 2 minutes, unscripted

The part you can’t outsource.

Three prompts, published when the build phase opens. Record in one take. Judges are scoring whether you understand what you made — polish is not the point.

  1. 01What did you build, and what did AI build?
  2. 02What was the hardest decision your team made, and why did you decide that way?
  3. 03What would you do next if you had another month?

Tooling

Which AI tool for which job.

All of these are permitted. None are required. Use whatever gets you to something real fastest.

Thinking and research
ChatGPT, Claude, Gemini
Pressure-test your idea. Ask it to argue against you, not for you.
Building without code
Lovable, Bolt, Replit AI
Describe what you want and iterate. Fastest route to a working demo.
Building with code
Cursor, GitHub Copilot, Replit AI
Learn what the code does as you go — you'll be asked to explain it.
Design and images
Midjourney, Flux, Figma
Good for mockups and brand. Don't let visuals outrun the idea.
Video and voice
Runway, ElevenLabs
Useful for the pitch video. Your own voice is usually more convincing.

One condition

You can use any of this. You just have to understand what you shipped — the founder Q&A is where that gets checked, and judges score understanding rather than output.
Registration closes November 1, 2026

Read enough. Go build.

Registration closes November 1, 2026. The build phase opens November 2, 2026 — everything above is waiting for you.

Register free How it works
  • Free — no entry or submission fee
  • Under a minute, six fields
  • No idea or team needed yet