Most "startup ideas" lists are useless. Vague categories, no examples, no numbers, nothing you could actually build on Monday.
This one is different. Greg Isenberg and Jonathan Courtney sat down and traded their nine favorite opportunities for 2026, with real companies, real revenue, and the exact angle to attack each one. I watched the whole thing and pulled out the best startup ideas worth your time, plus the framework that ties them together.
Read to the end. The last section, "date the product, marry the niche," is the part that decides whether any of these actually makes money.
1. Live, unscripted shows for high-value audiences
As AI sanitizes everything, people crave the opposite: messy, live, unscripted humans talking long-form. Jonathan pointed at Twitch culture, where streamers go live for hours and fans pay to hang out.
The business reason to care: TBPN, a live business show, sold for over $100M. Small viewership, extremely high-value viewership. The opening is obvious once you see it. What other high-value audiences need a live show?
You don't need scale. Jonathan gets 1,000 to 2,000 viewers each Tuesday and says that alone could make €400,000 to €500,000 a year through in-person events and merch, no ads. Patreon numbers for niche shows routinely hit five figures a month.
2. Action apps (agent-first software)
Most apps make you do the work. You scroll, you log in, you fill the form. Greg's favorite category flips that: apps that take action for you. Clear the inbox, book the calendar, file the expenses.
Think of the AI-first version of Superhuman that answers 80% of your email through agents and only surfaces the two messages that actually need you. The trick is reimagining the core experience, not bolting AI onto an old one.
You don't build the models from scratch. You build a layer on top of the Claude Agent SDK and similar tools, the same way "GPT wrappers" became a thing, except now it's agent wrappers. The winning formula: mobile-first, agent-powered, niche, with a customer who'll pay.
3. Solving loneliness (third spaces and community apps)
Almost 22% of Americans have fewer than one close friend. That's a crisis, and a category. Both hosts are building here on purpose.
The examples are concrete. Dads of Marathon, a Discord for casual gamer dads, hit 13,898 members. 222 runs a personality test and sends you to real dinners with five to seven matches in your city. Fabric runs 75-plus in-person gatherings a month and turned cheap commercial real estate into community space.
The build pattern: pick one niche, add a membership model for recurring revenue, and bring people together online or in real life. "Dads who play [game]." "Founders who [hobby]." The narrower the better.
4. Elder tech
Everyone builds for Gen Z and millennials. Meanwhile there are 70M-plus boomers in the US with money, free time, and almost nobody building for them. AI can help with hearing, mobility, memory, social connection, and vision.
One counterintuitive note: don't show gray hair and walkers in your marketing. These people don't see themselves that way. They just want products that make them happier and healthier.
Jonathan's own company, Facilitator.com, found its most valuable customers were over 45, not the 20-somethings they targeted. And Facebook ads still convert here, because that's where this audience actually is. Fish where the fish are.
5. Adult hobbies and learning for joy
Pottery, woodworking, drawing, painting, gardening, book clubs. As people push back against screens, paid in-person creative experiences are booming, and almost nobody wants to run them, which is exactly why the lane is open.
Jonathan ran a painting retreat that sold out instantly. A Berlin book club draws 85 paying people every Sunday. Dr. Tyler Lemco's Creative Club hosts paper-lamp workshops people pay to attend instead of going to a bar.
Run a free version and a premium version. The free meetup builds the audience. The high-end retreat, priced at thousands, monetizes it.
6. AI employees (verticalized digital workers)
Selling $5,000-a-month AI employees into businesses is one of the most obvious plays of 2026. The catch is positioning. Founders don't trust agents to replace a senior hire, but every company needs junior work done.
So sell juniors, not replacements. Pick one vertical and one job title, list every job-to-be-done for that role, and have your agents do them at a tenth of the cost. Start with two or three tasks, then add more until it's a real digital employee.
The companies that win here make the use case crystal clear: this agent does X, Y, and Z, removes this many hours from your week, costs a fraction of a hire, and switches on today.
7. Personalized nutrition from your own health data
For the first time, millions of people are testing their own blood work and gut biome, then dumping the results into AI to make sense of them. Function Health, ZOE, and Cal AI (around $50M ARR) are early proof.
The opportunity is to go vertical. Not generic health. ZOE for one specific problem. Roughly 60 million Americans, about 20% of adults, have GERD, and Jonathan says he'd pay almost anything for a single source of truth on it.
A simple way to find your vertical: walk into a CVS and look at which over-the-counter shelves are the biggest. That's where the pain, and the money, already is.
8. Pet health
The pet industry is worth roughly $140B, and under 2% of pets have any smart monitoring. People treat their pets like family now, so the willingness to spend is there.
The recipe is to copy human health and wellness, then point it at pets. "Athletic Greens for pets." "Eight Sleep for pets." Add a cheap AI brain to existing pet hardware like collars that track heart rate, sleep, and activity. PetPace and Whistle are early movers, which means it's still wide open.
9. AI-native media companies
The opposite of idea one, and some people will hate it. Use AI to build high-quality, clearly-disclosed media accounts, grow an audience, then sell that audience products.
Rowan Cheung grew an account to nearly 400,000 followers in 18 to 24 months using an AI avatar covering news. Slop won't work in 2026, but a genuinely top 1% AI-native media company in a tight niche, ideally with a human in the loop, can build an audience fast and monetize hard.
The framework that makes any of these work
Greg's rule: date the product, marry the niche. Most founders pick the wrong niche, build a great product, and still struggle, because the audience has no real pain or no money.
So choose a niche on three filters: one you genuinely care about, one where you have an unfair advantage, and one with both a sharp pain point and disposable income. The product can change as you learn it. The niche shouldn't.
Jonathan's proof: his painting retreat made $90,000 to $110,000 while similar retreats make $5,000, for one reason. He ran it for entrepreneurs, not the general public. Picking who you serve often matters more than what you build. Once you've married the niche, growth gets easier too, because everything from your content to your follow-up can speak to one specific person.
I break down founder playbooks every week → Watch the Podcast
Frequently Asked Questions
What are the best startup ideas for 2026?
The strongest categories right now are agent-first "action apps," AI employees sold by vertical, software that solves loneliness through community, elder tech for the 65-plus market, personalized health from blood and biome data, pet health, adult-hobby experiences, live unscripted shows, and high-quality AI-native media. Each works best when you niche down hard.
What is an "action app"?
An action app is software that does tasks on your behalf using AI agents, instead of making you do them manually. Examples include an email app that answers 80% of your messages automatically, or an agent that books your calendar and files expenses. You typically build it on top of an agent SDK rather than training models yourself.
Do I need a brand-new idea to build a startup?
No. Most of these opportunities are about taking a proven model and verticalizing it for a specific audience: ZOE for one health condition, Athletic Greens for pets, a community for one niche. The edge comes from picking a niche with real pain and real money, not from inventing something nobody has seen.
How do I pick the right niche for my startup?
Use the "date the product, marry the niche" test. Choose a niche you care about, have an unfair advantage in, and that has both a big pain point and disposable income. A painting retreat for entrepreneurs outearned identical retreats roughly twentyfold purely because of who it targeted.
None of these ideas are secret. That's the point. The hosts gave away the categories, the examples, and the numbers, because execution and niche are what actually matter. Pick the one you'd genuinely enjoy building, marry a niche with money and pain, and start this week.
