AI made it cheap for a small team to offer more. Most teams are spending that on offering what someone else already offers.
I should say up front that I do not know how this ends. Nobody does, and the next few years could prove this post wrong. What follows is a picture I cannot shake, and an article I read at a hard time that keeps attaching itself to it.
Dots on a map
Picture the economy of small companies and independent builders as dots on a map. Each dot has a radius around it: everything it offers. A company with a hundred products or services covers more ground than a company with one.
With AI, execution got cheaper in whole areas at once, and the radii are growing everywhere. Teams now build the adjacent product, the second service line, the feature a neighbour is known for, because it costs a fraction of what it used to. I understand the pull. When the cost of saying yes collapses, saying yes starts to feel like strategy.
The graveyard
In 2022, while I was caring for my father, I read an article about Google. He passed away later that year. The article’s argument stayed with me.
Its point was that Google’s vast resources were not helping it make products that last. The evidence is public. Killed by Google lists over three hundred products the company launched and later shut down: Google+, Google Reader, Inbox, Stadia, and hundreds more. Most were hyped at launch and used by real people. Some were killed while their users were still there. Each was staffed and funded at a scale a barely known startup never sees, and in the same years barely known startups shipped products that won.
The article traced it to where each product started. Google started from what it had: engineers, infrastructure, money, distribution. It shaped those into offerings and then looked for people to use them. The startups that beat it started from one need, with absolute focus, and scrambled for whatever resources that need required. The first way produces products that fit the company. The second produces products that fit a customer.
An essay from 2021 makes the same case, and quotes Waze’s former CEO on his years inside Google: time went to work that created no user value, which changed the company “from customer-focused to corporate guidelines focused.”
Widening is the Google strategy at small scale
AI gave every small team a little of Google’s position: spare capacity looking for somewhere to go. Expanding the radius because you now can is resource-first strategy. It starts from what the team is able to build and goes looking for a reason.
There is one difference, and it is the one that matters. Google funded three hundred burials out of a search business that never stopped paying. A small team cannot afford three.
More overlap, no new ground
The second thing I notice is where the radii grow. Almost all of it is into ground that is already covered: problems someone solved years ago, now solved again by a team that added them this quarter. The circles pile onto each other. The overlap on the map grows much faster than the area it covers.
Overlap is where competition turns into price and noise, and where a team’s reason to be chosen gets thinner with every offering that sounds like the neighbour’s. Very few dots are moving out into blank space, offering something that was not possible before. That was always the rare move. Cheap execution was supposed to make it less rare, and so far I mostly see it spent the other way.
I wrote once that when distribution got solved, finding what people actually want became the hard part. AI solved a second thing, the building, and left that hard part exactly where it was. Michael Porter put the old version of this in one sentence: “The essence of strategy is choosing what not to do.”
Cheaper execution is a reason to go deeper, and most teams are spending it on going wider.
The grand version of going deeper is a company with a reason to exist that nobody can copy by prompting. The ordinary version is three habits when the next offering comes up:
- Name one problem – Write down the one problem your team solves better than anyone within reach, because every new offering either sharpens that sentence or blurs it.
- Count the neighbours – List who already solves the thing you are about to add, because entering covered ground means competing on price with teams that got there first.
- Spend it inward – Put the time AI saved you into making the one thing better, because depth is what the next team with the same tools cannot catch up to in a weekend.
It is the same shift I see in code: when implementation gets cheap, the value moves into the decisions, and deciding what to build is the largest of them.
The case against me
The strongest argument against this post is that breadth sometimes wins. Parker Conrad built Rippling as what he calls a compound startup, one that deliberately builds “multiple different products in parallel” to solve problems that span systems. Sam Altman has said his group of tech CEOs keeps a betting pool on the first one-person billion-dollar company, something he calls “unimaginable without AI.” If either of them is right, the wide teams I am worried about are early rather than wrong.
I would correct myself this far. Breadth that serves one customer’s whole problem, one system instead of five, is still focus, measured at the customer. What I am worried about is breadth that serves the team’s capacity. Paul Graham’s advice to founders was to recruit users manually, one at a time, close enough to see the need. Being that close to a customer is still the part AI does not do for you.
Some years from now, the map will show which dots grew and which moved. I expect the ones still worth finding to be the ones that moved somewhere nobody else was.
Reach can be generated; a reason to be chosen has to be found. What did your team add this year that nobody within reach already offers?