In Plain Sight - 14
In Plain Sight is a weekly blog from Rhys that we will publish every Thursday/Friday. Each edition will be focused on a thorny topic within investing, startups, learning and work and policy.
In Plain Sight is partially a personal attempt to think independently and to avoid leaning too much on AI for answers to our questions. Will we use AI to edit and polish the text? Yes. But more importantly, will we come up with all of the ideas and analysis? Yes.
Here follows Edition 14 - 'Why validating a startup idea in 2026 is harder than ever'
------
For years, the validation playbook barely changed. Talk to ten customers, ship a landing page, run a survey and watch the waitlist fill. If enough people say yes, build it. A generation of founders learned from Paul Graham's "Do things that don't scale" that the fastest way to de-risk an idea is to go and ask.
We still believe in that instinct. But two things have changed:
- Building got radically cheaper. AI tools mean a working MVP takes days, not months.
- Honest signal got radically harder to find. Channels are noisier, signals are cheaper to fake and the people you're asking are more distracted than ever.
This bites hardest in consumer products, where the scarce resource is attention. B2B has its own version, which we come to below.
In 1971, Herbert Simon warned that a wealth of information creates a poverty of attention. In 2026, that poverty is acute. More than half of consumers say they see AI-generated "slop" in their feeds often or very often and Gloria Mark's research puts the average time we spend on one screen before switching at around 47 seconds.
Against that backdrop, "would people be interested in this?" is close to meaningless. Everyone is polite, distracted and drowning in things claiming to be interesting.
The better question is blunter: can you get ten strangers to give you fifteen focused minutes? If you can't, that may not be a distribution problem to solve later. It may be the validation result.
B2B isn't immune. Plenty of companies post impressive pilot-driven ARR growth, then watch it fall away once the first three months are up. A pilot is attention on loan. A renewal is attention earned.
The classic signals (survey responses, waitlist signups, "I'd definitely use this") were never perfect, but they used to be honest enough to trust directionally. Now they're trivial to fake. Cloudflare data shows bots account for 57% of requests to web pages, and 94% of login attempts are automated. Waitlists get inflated by referral farming. Surveys get filled in by people, sometimes with AI's help, who say something encouraging because it costs them nothing.
So there are now two kinds of yes:
- The cheap yes costs nothing: a signup, a like, a survey tick.
- The expensive yes costs something: a pre-payment, ripping out a tool they already rely on, a message asking when they can start.
When the cheap yes can be manufactured at scale, the expensive one is the only kind worth trusting.
There can be even tougher pills to take... You can build something that works, that does exactly what it set out to do and still find that no one downloads it.
That's often truest for products that are genuinely good for people. In an attention economy this steep, the most useful messages are often the hardest to land. Take exercise. The NHS guideline is 150 minutes of moderate activity a week: roughly a brisk 30-minute walk, five days a week. Yet the fitness content that wins attention is all 5am cold plunges, "dog mentality" and "who's gonna carry the boats?" We add drama to something that should feel normal and doable, because drama is what gets seen.
For founders in health, learning or financial wellbeing, this is the uncomfortable truth: proving your product works is not the same as proving people will choose it. Validation has to test both. Sometimes the real product is the framing that makes something sensible feel worth paying attention to.
So, if broad channels are the noisiest place to look, the best signal is probably with people who already feel the problem badly enough to talk about it somewhere specific: a niche community, a group chat, a forum, a founder's own network.
That isn't new. It's what "talk to users" always meant at its best. What's changed is the alternative. Broad, cold validation used to be noisy. Now it's barely signal at all.
The instinct might be to slow down and research harder. We'd argue the opposite. Because building is cheap, founders don't need to bet everything on one idea. They can build several scrappy, narrow takes on the same problem and see which one a small group can't stop using. Look at fitness, diet or tutoring: products solving the same problem through different mechanics, brands and communities.
This isn't an argument for staying shallow. When anyone can build, execution alone stops being an advantage, and the winners are the founders who understand their customer better than anyone else. But depth is expensive, so earn it. Go wide before you know and the go deep once you do.
None of this is a clean formula.
- Friction can kill frictionless products. For mass-market products where adoption should feel effortless, demanding an expensive yes too early can kill a good idea.
- Warm can become an echo chamber. The small community that loves an idea may not represent the market it eventually needs to win.
- Many bets spread you thin. A founder's own attention is their scarcest resource. Run too many experiments and you validate everything a little and understand nothing well.
So what would we advise?
For founders: treat earned attention as a headline metric. Ask early users for money, time or a real switching cost, not just a click. Find the small, warm community before the big, cold audience. Run several small bets before committing to one.
For angels, accelerators and advisers: "Have you talked to customers?" and "How big is the waitlist?" need an update. The better question in 2026 is "what did it cost the people who said yes?"
Validation has never taken more judgement. That's good news for founders willing to do the harder, more honest version, because most of their competitors won't.
If you're wrestling with how to find honest signal in a loud market, or building something that helps others do it, we'd love to talk.
