Since my genies seems to have all gone to rehab at the same time I have leisure (!?) to write this.

In the Expand phase, growth isn’t a curve. it’s a staircase. You grow until you approach a ceiling—some rate-limiting resource.

Either increase the supply of that resource or reduce the consumption until you get back to growth. Disaster averted!

Then the next rate-limiting resource looms and you repeat it. Then the next one. Then the next.

Eventually you know (from bumping into them) what the rate-limiting resources are and how to keep the supply curve above the demand curve. Then you can shift into Extract mode.

(Successful companies often retcon all these near-death Expand experiences to make their success seem inevitable.)

Genies

Everybody cuts limits at once. Not a coincidence, a signal.

The genie is in Expand. Hard. Usage is growing faster than almost any product in history.

When you’re about to hit the ceiling, you have two choices. To bend the supply curve up—build more data centers, get more chips, make inference cheaper. To bend the demand curve down—slow the growth, ration the usage, make free/cheap tiers less attractive.

The model providers are doing both, but bending demand down is the surprise move, at least to me. Who shuts the rocket motor off mid-air?

The twist here is competitive dynamics. Normally, bending demand down in the face of competition is suicide. Users leave. They go to the competitor. You lose. Expand doesn’t forgive.

The Bottleneck?

In Expand, the first question is always: what’s the next rate-limiting resource?

Chips? Nvidia supply is constrained, H100s are scarce, everyone’s fighting for allocation. Except Google makes their own. Amazon makes their own. Anthropic has a preferential supply agreement. And all three still cut limits at the same time. Not chips.

Raw compute capacity? Data centers, cooling, power delivery. Real constraints, multi-year buildouts. But physical constraints hit different companies at different times — different footprints, different geographies. You’d see variation. You see synchrony. Not physical capacity.

The economics of inference? The ratio of what it costs to serve a query to what users pay. Broken at scale, especially for free users. This one feels right until—they have basically unlimited capital. The compute bill is large but fundable. They’re not cutting limits because they literally can’t afford to serve you.

So what is it?

The story. Specifically, how long investors will fund giving away expensive capability while waiting for profits to catch up. That story has a shelf life. At some point you have to demonstrate a path to profitability, not just assert one. Usage limits are evidence you’re managing toward that—it’s a signal to investors, not a “we’re running out of money” decision.

That’s why all three moved together. The same investor class, the same stage, the same moment when “trust us, it’ll work out” stops being enough.

The bottleneck isn’t engineering. It’s narrative.

It’s a Race

What breaks the model cartel? Someone bends the supply curve up. Fixes the narrative. Inference gets cheaper through distillation, caching, smarter routing to smaller models. Custom silicon matures. New data center capacity comes online. One company gets meaningfully ahead on unit economics and can afford to open the throttle while competitors can’t.

That company wins the next wave.

What’s Next?

I write about augmented coding—developers working with genies all day. Usage limits bite differently for us. A power user hitting a daily cap mid-flow isn’t mildly inconvenienced. Their work stops. They have to, I don’t know, write another blog post.

For developers, caps pressure them toward the API—explicit pricing, higher ceilings, no daily cliff. For everyone else, it’s just a wall.

Limits split the user base: casual users get free but capped, developers get metered but uncapped, and the middle—technical-but-not-API-savvy power users—gets squeezed into paid consumer tiers. That’s the conversion pressure the companies actually want, even if a bit more revenue isn’t going to change the global constraint.

Uncomfortable question: are the limits temporary—a bridge while supply catches up—or are they the beginning of a new equilibrium where heavy AI usage is a premium product, not a default?

In 3X terms: is this a brief pause on the way to the next Expand staircase, or the beginning of Extract—where growth slows, margins matter, and rationing becomes a feature?

I don’t know. But I’m watching which company bends supply up first. That tells you everything about where Expand goes next.

Can we use Skills to get the genie to work TCR-style? (All the coolest ideas are blends of 2 or more existing ideas.)

Thank you , , , , , and many others for tuning into my live video! Join me for my next live video in the app.

Get more from Kent Beck in the Substack app

Available for iOS and Android

Get the app

Transformation: Map the effects & how they affect each other. Find the cycle with an even number of inhibitions. Go upstream & push one of the effects the opposite direction.

Okay, that’s an abstract brainful. But this is the most powerful tool I have for making big changes with small efforts, so it’s worth it. I promise. We’ll start with an example.

Read more

Read more

Brought to you in partnership with WorkOS and Augment Code.

Why can’t tree-huggers & forest destroyers get along?

I lived in the wilds of southern Oregon for 17 years. I enjoyed many things about life in the forest. One of the things I loved most was watching the politics. Our community was polarized before polarization was cool. Loggers versus environmentalists, sure, but also weird alliances. Open-carry fisherman cozying up to tree huggers to save the salmon & steelhead. Rugged loggers cozying up to (relatively) soft townies with money.

My roots went deep in the area, one set of grandparents having moved there in 1943. The other grandparents (that grandfather was a former mayor & city councilor) moved there in 1933. I felt connected & invested, even if I didn’t usually stick my ideas out there.

Gridlock

The one time I did put my 2 cents in (this is back when we had cents) was when I tried to resolve the absolute hatred between environmentalists & loggers using what I was studying about incentive structures. I think the system I came up with was cool but it quickly disappeared, so now I want to put it in public permanently here.

Here’s the setup:

• Only 5% of the old-growth forest was left.

• Loggers wanted to harvest all of it.

• Environmentalists wanted all logging to stop.

• Second-growth forest was prone to catastrophic wild fires. (There’s a special kind of helpless feeling watching a fire approach your home.)

• Loggers wanted to harvest second-growth.

• Environmentalists wanted all logging to stop.

The result was a complete impasse. Forests burning. Mills closing. Crime & drugs up. Anybody with any ambition leaving. Nobody was getting what they wanted—loggers, environmentalists, workers.

(Or at least what they said they wanted—there seemed to be a bunch of psychodramas playing out.)

Incentives

I recast the forest tinning situation as an incentives problem (I was intensively studying incentives at the time). Once the loggers finally got permission to harvest a tract of second-growth, they were incentivized to take out every stick of wood with economic value, leaving further growth stunted, encouraging the growth of flammable underbrush.

Because environmentalists saw the loggers’ incentives, they were ever more incentivized to block all logging & put onerous restrictions on any activity that managed to sneak through. In Influence Diagram terms, more logging leads to more money & more damage, but more damage leads to more resistance leads to less logging.

Classic inhibiting loop. More logging leads to less logging. (We could go on & on mapping this system, but this will do to illustrate my pr of the many interventions we can make in a system is to speed or slow feedback. What if, instead of getting paid for this harvest, the loggers got paid for the next harvest. Today they’d thin the forest, with any valuable material going to be turned into products, but it wasn’t until 10 years later that they would be paid the proceeds of the next forest thinning. You would get paid more if the forest thrived over the next 10 years, less if it grew less.

(I think I kind of made up the notation for delay.)

Now we have a reinforcing loop. More logging. Less damage (because the loggers get paid in a decade). Less resistance. More logging (in the form of forest thinning.)

Priming the Pump

That first logger, how do they get paid? They are paying for diesel, salaries, depreciation today & won’t get money for 10 years. The right to be paid in 10 years can be turned into a financial instrument to be sold today (remember those soft townies with money?) Now you have monied interests who also care about the forest’s health.

And who is best suited to evaluate the health of the forest for future gain (and avoiding future loss from pests or fire)? Well, those environmentalists who care so about the forest are well positioned to act as consultants & auditors.

Mill workers would be back at work. Local capital would have another way to extract rents. Environmentalists would have healthier forests. Loggers would have trees to cut.

Chickening Out

I wrote the above up as a letter to the editor of the local newspaper. That got me invited to a “summit” of conservationists & loggers. When the microphone got around to me I had an attack of shyness, said something self-deprecating, and passed the mic on to the next person. So that was that.

Would it have worked? Maybe. Entrenched interests were more interested in staying entrenched than in making progress. That’s true today in many situations I see. It’s not as simple as “change the rules and the behavior will change”. But “don’t change the rules & the behavior will definitely not change”.

Kent partners with a handful of companies each year on editorial collaborations, speaking, and workshops. If that's interesting, let's talk →

That was one of the insights behind XP—fix quality at “high”, let scope vary, & you’ll end up with sooner & cheaper. Seems like magic to break the Iron Triangle, but it’s not.

Read more

The first thing we finish will earn the resources for the next thing which will earn the resources for the next.

Different Game, Different Rules

Here’s The Finish Line Game, a different software development game:

We want software that does X. Once we’ve reached X we’re done. It happens. I have some data. I need it munged. I write a munging script. Finito. I’ll never use that script again.

The hidden assumption behind this style of spec-driven development is that we’re playing The Finish Line Game. Get the spec right. Get the desired software. Done. No need to consider the future because there won’t be one.

Failure & Success in the Finish Line Game

If you’re playing The Finish Line Game, you can still fail. Design can be so bad you don’t cross the finish line. Tests can be so bad that you don’t notice that you haven’t crossed the finish line. The “finish line” itself can shift without you noticing.

In the Tidy First world, we illustrate by contrasting “futures” & “features”. Futures are what all you can implement next. The genie is no good at managing futures. If you’re playing The Finish Line Game, you’re betting that features will cross the finish line before futures run out.

This is “better agent.md file” territory. If we leave the genie autonomous, can we nudge its work to be good enough not to fail? Sometimes yes, sometimes no (in my experience). And the genie keeps getting better. But it’s still the same Finish Line Game.

Playing The Compounding Game

Just as football tactics won’t work on a baseball field, spec-driven development doesn’t work if you’re playing The Compounding Game. A better spec will never get you from dollar sign N to dollar sign N+1 forever. A better agent file won’t extend the lifespan of the system long enough to get from N to N+1. At some point the complexity will exceed the genie’s capacity & it’s game over with lots of dollar signs left to play for.

Instead, playing The Compounding Game requires as much investment in futures as in features. The two alternate. (Hence the “tidy first” connection—tidying is part of investing in the future.)

We’ll talk more about those vertical shifts more as we go along:

• What senior engineers do to shift vertically

• What junior engineers do to shift vertically

• What tools to write

• What practices to apply

• How to get the genie to help

I’m writing this in a creative storm, so it’s not polished. Thought it was worth getting out to y’all & getting your feedback.

The “natural” use of genies is to further isolate people who are already suffering (even if they enjoy it) from isolation. “Teams of one can now achieve infinite results!!!” Technical solution to a human problem. (There should be…

Read more

People plan & God laughs.

As last year moved into its second half, it became clear that the material in Tidy Together was being overtaken by events. The chan…

Read more