Good morning everyone, happy Friday, and before we lean in here, are you sitting up straight?

Odd question to ask you as I lie down typing this, but an experiment result was published this week showing that if you sit up right, you make better risk-taking decisions, will likely be in a better mood, and will feel more proud. TLDR, researchers manipulated how people sat by controlling minor things on their desk setup and tested them on various things. Nobody was told to sit differently. They tweaked the environment slightly, and the behavior followed.

Which if you think about it, is kinda the entire point of product design → your users’ posture inside your product is whatever your defaults make it. Just a thought!

And a quick note before we get going: I started making videos this week. Short, one-take, over-my-morning-coffee versions of these type of ideas. If you enjoy my stuff and want to see it, feel free to subscribe!

For example, “Why does everyone want to work with Christopher Nolan?” shares some bits on what he does on set that make him so remarkable, and what we can learn from that of course.

In the next 5 minutes over your coffee: the small detail behind why these 90-second videos print cash, proof that your roadmap might be making you worse at everything on it, Spotify’s plan to sell fans their own remixes, the subscription moat that might be disappearing, and the $450M round that financed customers instead of equity. Sit up straight for this one.

Guys do you remember Quibi? I wrote their fail story a while. They raised billions super fast and died even faster. Well, ReelShort is a near identical app doing 90-second, mobile-native soap opera episodes, and this has been their revenue track: $97M in 2023 → $400M 2024→ $785M 2025 → a billion now. The micro-drama market outside China is $3.6B this year, projected $9.5B by 2031.

So why ReelShort and not Quibi, given it’s the exact same idea—really bad, short-form episodic video just fire hosed at users though an app? The trick is the payment model. Quibi, like all other major streaming apps, charged a subscription before you cared about the content. ReelShort gives you 10–20 episodes free, walk you to the cliffhanger where the CEO is about to discover the twins are his, and then asks for coins to continue. No commitment, no monthly anything. They’re betting on micropayment at the precise moment people are most interested in what happens next. Even for bad content, the paywall placement is world class. Where you put the ask matters more than what you ask for.

Go find your product’s cliffhanger. What is your moment where user motivation/desire/necessity is provably the highest (e.g the almost-finished report to download, the locked insight), and look at what your paywall is doing there. Most ask at signup when desire is at its lowest. The MPA breakdown of the category is interesting if the economics interest you. Otherwise this is the UX breakdown 👇

If you read Tuesday’s edition, you know I went deep on the ambient AI play, and Attio is the company running it best. Fields in your CRM that research and fill themselves, sales calls that become structured data on the right record on autopilot, a CRM that built itself from my inbox in minutes. It’s how I run my whole sponsorship business as a one-man shop.

Missed the deep dive? Catch up here. Or skip ahead and just try the thing…

A new study from Berkley says that switching between goals—not tasks—without a doubt degrades performance on each one. People did the same cognitive task while the objective flipped between “be fast” and “be accurate.” Same task. Just a different goal. Every flip cost performance while the brain reconfigured, and the less transition time people got, the worse they did.

Yes we all know task-switching is bad. But this is sneakier: the task never changed, only the definition of good changed, and that alone made people worse. Which is what a startup does to itself every other week—this sprint is growth, no actually retention, nvm now it’s enterprise. Same team, same work, different goal. But it’s not just startups obviously. I find myself doing this personally a fair amount too.

Try hold onto just one definition of success per cycle, and no mid-week goal flips. Remind yourself that your WIP limits are not for how much you can get done, but for the objectives behind why you’re doing them. The study summary is a 4-minute read and mildly incriminating for a lot of us.

If switching goals makes you worse at all of them. The fix might not have to just be discipline, but delegation—handing off whole tasks and goals instead of splitting your brain. That’s where Tendem can help you: just describe the work you need done right inside Claude or GPT, see a price, approve it, and one of 10,000+ vetted human experts send it back, finished and accurate, while you stay on your own goal/task.

It’s the first and only platform I’ve seen doing the You → Your AI → Expert Human → back to You loop. It genuinely works and beats pure AI output.

Taylor Swift and Kanye West. Drake and Kendrick. If you pay for Spotify, you might be able to squash the beef and make up your own song of the summer with them. Made with AI. And legally. What!?

The deal was back in May between Spotify and Universal, and this week’s earnings put it back in the spotlight. The mechanics are interesting part: opt-in only and every single play of a fan-made remix pays the original artist and the songwriter separately, on top of normal streaming royalties. Spotify’s three words for it: consent, credit, compensation. No price or launch set yet. US first.

And if you’re wondering whether anyone actually wants this: the viral AI ’60s soul version of 50 Cent’s “Many Men” is exactly the demand this deal endorses. To be fair, this was fire.

The music industry has been mega anti AI, and fairly so. Their answer has just been lawsuit, lawsuit, lawsuit. So, this is a big deal because is the first real reply the other way. Basically, if AI is going to raid your catalog anyway, mind as well sell first class tickets and make money from it. Pretty much the same thing as the early 2000s shtick when Spotify told studios to join streaming, or get eaten by Napster. In this new world, the fans become a licensed content factory, artists get paid per generation, and Spotify gets a new ARPU lever on its Premium base without producing a single new song. Who do you think wins the training-data fight by simply routing around it with a product? 🧐

Join it vs try beat it. If AI is going to eat your category’s content, a countermove is a licensed product. Ask what your users would make with your assets if they were allowed to—that’s a new SKU waiting to exist. For example, I LOVE how Lenny gave his readers all his transcript data from podcasts and called on the community to go make something fun with it. Like this RPG game…

The average American pays $219/month for subscriptions and, when asked, guesses $86. That’s a 89% underestimate of what they are spending. And I can confirm first-hand (my wife has been using an app for this recently and culling things), that the behavior is shifting and it matters. People are cutting the number of subs they have (subs per person are down -6.4% YoY), with 47% having canceled at least one paid service in the past six months.

That $133 gap between reality and perception has been one of the most profitable moats in software—the subscription amnesia lots of companies have enjoyed being part of. Except the tailwinds are clearing the fog for people with banks doing more to flag recurrings, money apps like the one Julia uses that go hunt them for sport and allow you to pay what you think is fair, and the obvious next thing is AI agents that read the statement and do the canceling themselves. “They forgot about us” is not durable revenue anymore. When the auditor comes, the only subscriptions that survive are the ones that can defend themselves in the two seconds before the cancel tap.

Design for the audit day on purpose. Show a value receipt (“you used this 14 times this month”). Offer pause instead of cancel—65% of subscribers say flexibility is the #1 reason they stay. And don’t bury your cancel flow.. The full numbers are here and they’re interesting.

Function Health—the whole-body testing membership—just took $450M in growth financing from General Catalyst. This is just 8 months after a $298M Series B. But this isn’t a priced round, rather it’s purely for financing customer acquisition directly, no equity given out, and gets repaid from the revenue those customers generate. Function keeps its cap table. GC buys a piece of future cohort cash.

This deal works because Function’s math is provable: 160+ biomarkers per member, a subscription with known retention, and they just bought Getlabs (a nationwide blood-draw network) so they even own fulfillment. AKA, if you have hard numbers on your CAC, LTV, and payback period, your acquisition spend could become a financeable asset, like inventory or receivables. In this case, equity built the machine. And this lent capital feeds a machine.

You don’t need to be huge, rather big enough so your math is legible. Clean cohort tables, real payback windows, churn you can defend. Even at small scale that data is leverage—with lenders, sponsors, partners, and eventually funds like this one. If you can’t produce your cohort math, that’s a good starting point. Why GC structured it this way is a good read.

I’m being naughty and taking over the fun corner today to ask you to check out my app, Good Times.

I made it because social today sucks. Ads everywhere. AI slop. Algos feeding us. Performance for strangers. And I find group chats go quiet. It’s just not easy to stay connected with your favorite people in a way that feels fun, light, and actually meaningful.

Good Times is one question a day with your friends. No AI. No Algorithms. No Ads. No Strangers. 100% free, private group social.

On iOS only right now, Android in closed testing.

Try it out. Make a group. Answer today's question, and please let me know what you think if you do!