151 experiments · one from every episode

Founder Experiments

Every episode ends with a hands-on experiment you can run this week. No theory. Just build.

Try:

Get a new experiment every week

Each episode's hands-on experiment, the tools to run it, and the AI news founders need. One email, free.

No spam. Unsubscribe anytime.

What do you want to try?

Pick an area to see the experiments for it.

Random experiment generator

Not sure where to start?

Hit the button and we’ll hand you one of 151 experiments to run this week.

All experiments151

The 48-Hour Shadow AI Count

  1. 1

    Inspired by Donella Cohen and James Auchterlonie of MLNavigator. Give it 48 hours. Ask every person on your team one anonymous question: “Which AI tools did you use for work this week, and was it a company account or your personal one?”

  2. 2

    Tally the answers against the list of tools you think your company uses. James and Donella found that companies often have AI policies with no idea whether anyone follows them. The gap between your list and theirs is your real risk surface.

  3. 3

    Run the Shadow AI Agent Inventory Builder skill on the results and fix your top gap.

Stretch goal: Repeat the count in 90 days and see whether the gap between your list and your team’s list got smaller.

The One-Page Manifest Test

  1. 1

    Inspired by Jason Katz of Kindling Solutions. For the next two weeks, every internal AI app, automation or agent your team wants to build has to pass through a one-page manifest first: what data it reads and writes, who uses it, what it can do without a human, its monthly spend cap and who can shut it off.

  2. 2

    Load the Microapp Manifest Builder skill into Claude or ChatGPT. Ask your team to run every new build request through it and send the manifest to one approver. Approve or reject within 24 hours.

  3. 3

    Measure how many requests came in, how many duplicated data that already existed, how many asked for more access than they needed, and how long approval took.

Stretch goal: What you’ll probably learn: your team is building more than you thought. The manifest doesn’t slow them down. It shows you which builds deserve a real company brain underneath them.

The Five Steps Ahead Week

  1. 1

    Inspired by Brittany Michalchuk’s habit of asking what today’s action costs you in the future, before you take it. Before you start any new tool, subscription, meeting series, or project this week, write down its five-step consequence chain, from this week out to 12 months.

  2. 2

    Then decide: make the move as planned, reshape it into something smaller, or drop it. The Five Steps Ahead Decision Filter can run the chain for you.

  3. 3

    At the end of the week, count how many moves you made, reshaped, or dropped.

Stretch goal: Brittany’s question for every move: what does this action cost you five steps from now?

The Walk-Away Test

  1. 1

    Inspired by James Hurlock of Fitura Brands. Before you sign your next lease, vendor contract or big partnership, write down three conditions that would make you walk away. Set them before you see the final terms.

  2. 2

    Then run the Lease Walk-Away Scorecard, or score the deal by hand.

  3. 3

    If any walk-away condition gets triggered, sleep on it for 48 hours before you sign.

Stretch goal: James’s rule: you can fix almost anything after you open, except a bad lease.

Sponsors

CEOfriendInbox AlchemyHowdiAINative StudentSnipdMomentous

The Context Passport Test

  1. 1

    Inspired by Adam Wright of WhatIff. Hypothesis: your context is worth more than your favorite model.

  2. 2

    Run the AI Context Passport Builder once. Paste your Work Pass into a model you’ve never used before, then give it the same three real tasks you’d normally give your go-to assistant.

  3. 3

    What to measure: How many follow-up questions it asks, how many minutes you spend re-explaining, and whether the output is good enough to ship.

  4. 4

    Run it for 7 days, switching models at least twice.

Stretch goal: What you learn: Whether you’re locked in to a model or just to your own missing documentation.

The 14-Day 10/20 Reset

  1. 1

    Inspired by Jonathan Berent of NextSense. For two weeks, take one afternoon rest break using the 10/20 method: eye mask on, rain sounds on a 10 minute timer, backup alarm at 20 minutes. Sleep is optional.

  2. 2

    If you wake at 3 AM and can’t drift off within 20 minutes, get up and spend that time on one project you keep postponing.

  3. 3

    Why it works: Quiet, eyes-closed rest pushes most people into a calm alpha rhythm within minutes, and capping naps at 20 minutes keeps you out of deep sleep and the grogginess that follows it.

  4. 4

    What to measure: Energy at 10 AM and 4 PM on a 1 to 10 scale, plus total minutes of deep work captured during early wake-ups.

Stretch goal: Success looks like: Higher afternoon energy scores in week two and at least one stalled project moving forward.

The One-Question Posting Challenge

  1. 1

    Inspired by Ariel Cohen of Swan. For the next 14 days, ask one question before every post: “Will my audience find this valuable?” If you can’t name the single takeaway a reader could act on today, don’t publish.

  2. 2

    Find three outlier posts in your niche and borrow their format, never their words.

  3. 3

    Keep a log of posts you shipped, posts you killed, and the comments, DMs, and profile views each shipped post brought in.

  4. 4

    Success metric: fewer posts, more conversations.

Stretch goal: Bonus round: message five people who engaged and see how many replies you get compared with cold outreach.

The Five-Interview Signal Test

  1. 1

    Inspired by Rickard Hansson of Gainable. Before you build, put on your research hat and interview five people who fit your buyer thesis. Don’t sell. Ask about their pain, then ask them to dream: if they had unlimited resources and nobody saying no, what would they build?

  2. 2

    Run it: Book five 30-minute calls with people in one specific role. Same questions for all five.

  3. 3

    Read the signal: If all five describe the same pain, stop interviewing. The sixth and the hundredth will say it too. If you get five different answers, your thesis is wrong. Rewrite it and find five new people.

  4. 4

    Watch for the surprise: Rickard found his buyers wanted internal apps but could not describe them in a prompt. That one insight became Gainable’s whole design.

Stretch goal: Time cost: About two weeks. Bonus: Interviewees who feel heard often become your first pilot customers.

The Zombie Test for Your Pitch

  1. 1

    Inspired by Pablo Starr of Brain Poison Studios. Hypothesis: your pitch or launch video loses people because the threat is abstract. Rewriting it so the danger is as obvious as “zombies are coming” will keep more people watching or reading to the end.

  2. 2

    Write your current pitch’s threat in one sentence. If you need the word “inefficiency” or “friction,” it fails the test.

  3. 3

    Rewrite it with a named enemy, a specific loss, and a deadline.

  4. 4

    Post both versions as a short video or LinkedIn post a week apart, or send each to half your email list.

  5. 5

    Compare completion rate, replies, and click-through.

Stretch goal: Time required: 90 minutes plus one week of data. Success metric: at least a 20% lift in completion or reply rate on the rewritten version.

The Unbox Your Own Product Test

  1. 1

    Inspired by Rafi Moshe of StoriedLife, who spent eight months polishing AI features before he ordered his own printed book. The moment he held it, he understood why customers were buying, and he rebuilt the homepage around the physical outcome. Hypothesis: your homepage sells the features you’re proud of instead of the result your customer actually wants to hold.

  2. 2

    This week, buy your own product with a personal card as a brand new customer. Go through the full journey, from ad or search to checkout to the first delivered result.

  3. 3

    Write down the single moment that made you feel something.

  4. 4

    Rewrite your hero headline and first image around that moment. Leave the feature list underneath.

Stretch goal: What to measure: homepage to signup conversion over the next 14 days against the previous 14. Time required: one afternoon, plus two weeks of data.

The Decide-Once Week

  1. 1

    Inspired by Gelu Sulugiuc of Financial News Systems, whose trading rule is to do your homework, write your thesis, and set your loss limit before the news hits. Hypothesis: writing if-then rules for one recurring decision before it happens will cut decision time and regret in half.

  2. 2

    Pick one decision you make at least weekly under pressure (discount requests, ad spend, candidate yes or no).

  3. 3

    Write 3 to 7 IF / THEN rules plus a hard loss limit before Monday. Setup takes about 30 minutes.

  4. 4

    For seven days, follow the rules exactly. Log every time you’re tempted to break one.

Stretch goal: What to measure: minutes spent per decision, number of rule breaks, and how you feel about the outcomes at week’s end.

The Weekly AI Open Mic

  1. 1

    Inspired by Joe Esteves of Maine Pointe, who found the old suggestion box sat empty. A weekly open mic turns bottom-up innovation into a habit and shows you who is running toward the mission.

  2. 2

    Block 30 minutes on the same day each week and give every person on your team the mic for a few minutes.

  3. 3

    No slides, just a screen share and a two-sentence story: “Here’s the problem, here’s what I built.”

  4. 4

    The founder goes last and only asks questions.

  5. 5

    Keep a running list of every idea shared.

Stretch goal: What to measure: how many people share each week, how many ideas make it into a production workflow within 30 days, and who keeps showing up with something new.

Thousands of founders said yes and kept going.

Our free newsletter explains what they're actually doing.

Check the Newsletter