In partnership with
For founders shipping real product with AI in the loop. An essay worth your time, plus the sharpest things I've read, filtered hard. No hype. Just signal.
⏱️ The 30-second version
Your product is too complicated for one segment of your customers and too thin for another, and both are paying you this month. The meeting that decides to simplify only ever hears from one of them.
Sharpest read this fortnight: “AI is making us build too much”, on why AI made everything free to produce and nothing cheaper to own.
This issue's question: pull up the screen where new customers stop coming back. Before you remove anything from it, can you name the one job most of them arrived to do? Vote below.
📄 The essay: “Your product is too complicated. And too simple. At the same time.”
Somebody in your company is going to say the product feels bloated, and everyone in the room is going to agree.
That meeting is where most product teams do their damage, and it feels wonderful. It feels like taste. Like finally saying the thing everyone had been too polite to say. Then a quarter of engineering goes into removing options, the product comes out measurably cleaner, and none of the numbers you funded it for move.
I have sat in that meeting plenty of times, usually arguing to cut. The question on the table is always some version of “is our product too complicated,” and I have come round to thinking that question is broken. It is missing its object. Too complicated for whom, doing which job, on which attempt.
The evidence everyone quotes for simplifying is a jam table from 2000 that nobody has reliably reproduced since. The evidence that actually moved a number was a hospital changing where it filed critical results, with the same quantity of information on the screen. Physician accuracy went from 68% to 89%, and every simplify meeting I have sat through would have scored that change as doing nothing.
📡 The radar
The best things I read this fortnight on building product with AI. I read them so you don’t have to. Click the ones that earn it.
Building product in the age of AI · 6 min
AI is making us build too much
By Simon Green, writing on AI, ownership and organisational complexity.
AI made every artefact almost free to produce and nothing cheaper to own. His line is the one to keep: “Tokens are getting cheaper. Cognitive load is not.” Reading Steve Yegge's account of a 50-agent system that grew a 600,000-line factory and appointed a Head of Wheelhouse Law to govern its own governance, Green lands the detail that stuck with me: the people using the product asked for the pace of change to slow down.
Craft & judgment · 5 min
The cost YAGNI was never about
By Kent Beck, creator of Extreme Programming and co-author of the Agile Manifesto.
“You aren't gonna need it” was never about saving typing effort. It is about the cost of speculative structure, which sends two bills: you spend the option to build the right thing once you know what it is, and you pull cost forward while pushing revenue back. Both survive free code generation untouched. Cheap generation “makes the violation cheaper to commit, which is worse.”
Building the business · 6 min
The silent churn: why your best customers leave without a word
By Jason Lemkin, founder of SaaStr, 2x founder with nine figures of exits.
Written from the customer's side, which is what makes it land. Lemkin's company paid a vendor $60,000 a year for eight years, quietly halved their usage, and nobody ever asked why. The part worth arguing about with your team: fewer support tickets is not good news. Sometimes it means customers have given up trying to make your product work.
The counter-argument · 12 min
Progressive disclosure: from training wheels to week-long AI agents
By Jakob Nielsen, 43 years in usability, co-founder of Nielsen Norman Group.
I am linking this because it argues against me, and it does it well. Nielsen walks the same 1984 training-wheels study my essay leans on and draws a different conclusion: progressive disclosure “splits tasks by frequency, not people by skill.” His workbench test is worth running regardless of who is right: can a first-time user complete your top task using only what is visible on first load?
🔧 One thing worth trying
A single thing I actually used this fortnight. For you, not a listicle.
Read the document against itself, not just against its sources.
What it's for: Catching the mistakes that survive fact-checking, because every fact in them is true.
What makes it click: A strategy document I wrote for a client opened on the market gap: the Kennel Club had closed its Assured Breeder Scheme, so there was no accreditation left and the platform could fill it. Every part of that is accurate and sourced. The client sells cats. The Kennel Club is dogs, the cat equivalent is still running, and the whole argument for why breeders would sign up rested on a body none of them belong to. It survived every review that checked claims against sources, because the claim was true. What caught it was reading the opening against the audience described forty pages later. Now I read a finished document once more asking a different question: not "is each part right" but "do these parts agree with each other."
When to skip it: Anything short enough to hold in your head at once. This is for documents long enough that no single reader is holding all of it, which is exactly when it stops happening by accident.
🚦 The builder is the worst judge
6 min
The rise and fall of Homo Logicus
Jeff Atwood, twenty years ago, on the one bias that never dates: “of all the professional hubris I've observed in software developers, perhaps the greatest sin of all is that we consider ourselves typical users.” He is blunt about what that makes you. “We're not even remotely average, we are the edge conditions.” Worth reading next to this issue's essay, because the physicians in that hospital study could not feel their own improvement either. The people inside a screen all day are the worst available instrument for judging what it costs everyone else.
Your turn
Pull up the screen where new customers stop coming back. Before you take anything off it, work out the one job most of them arrived to do, and ask whether that job has a place of its own on the screen. Then measure your two segments separately, because a blended activation number hides both of them from you.
Think about the screen where your new customers stop coming back. When your team last talked about it, the proposal was:
Know a founder who’d get value from this? Forward it.
Not subscribed yet? Get it in your inbox →
Build Less, Solve More is from Boald. We help founders build digital products that matter, with AI as the engine, not the pitch.
Everything GTM. One platform.
Small teams don't have time to stitch together five tools and hope it works.
Apollo gives you everything you need to find leads, reach them, and close deals — all in one place:
230M+ verified contacts
AI-powered outreach
Data enrichment
Inbound lead capture
Meeting scheduler
And more
Stop juggling tools and start building pipeline that scales.
With Apollo, the AI revenue engine powering 4M+ users.