Nobody Sells Glue for Real Legos

01. The Two Bins: It Holds Fine Until You Lift It

Put two bins on a table. One is Lego. The other is twenty brands of off-brand bricks from three different dollar stores.

Pick up a knockoff and look at it. It's fine. Studs are there, tubes are there, the color is a shade off but you'd have to hold it against the real one to know. Snap two together and they hold. Build a wall six inches high and you would swear you bought the real thing.

Now build a castle. Take an hour. Do it properly.

Then pick it up.

The walls part at the seams. The corners sag. Bricks pop out of the middle of a run and the west side comes down in your hands. Nothing you did was wrong. Every piece you used was a piece that fit.

02. The Defect Has No Author: Every Brick Is Well Made

Here is the part most people get wrong when they tell this story. They blame the bricks.

The bricks are fine.

A dollar store brick isn't made badly either. It is made accurately — to its own mold. It is correct. It is simply correct against a different reference than the brick beside it, and a hair of difference doesn't show in two pieces. It shows in forty. Error does not cancel out. It accumulates, and it accumulates in the direction of whatever load you eventually put on it.

Now say it in software.

Every plugin in your install was written by someone who did the work. Read the source sometime — most of it is careful. Tested, documented, handling the edge cases the author knew about. They did their due diligence and they were right.

They were right about the world as it existed the day they shipped.

Then the world moves once. PHP goes to 8. A core release changes a hook. An API deprecates a field. A security patch tightens a default. And forty independently correct things go subtly, simultaneously wrong — not because anyone made a mistake, but because they were all measured against a reference that quietly stopped existing.

That is the actual problem. Not that somebody screwed up. That nobody did.

A failure with an author has someone who can fix it. A failure that emerges from the seams between forty correct things has no author, which means no owner, no ticket, nobody to call, and no version of "fixed" that stays fixed. You open a support thread and every vendor is polite and every vendor is right and nothing gets repaired — because the defect isn't in any of the parts. It's in the gaps between them, and nobody on earth sells gaps.

03. Credit Where It's Owed: WordPress Was Genius

Worth saying plainly, because the rest of this doesn't work if I'm just being ungenerous.

WordPress was a genius idea, and it was exactly the right idea for the moment it arrived. In 2005, if you wanted a website you could update yourself without ringing a developer every time your hours changed, your options were bad and expensive. WordPress said: here's the structure, it's free, anyone can extend it. Millions of people got a business online who otherwise would not have. That is one of the real democratizations in the history of the thing.

And it decayed. Not through malice, and not really through neglect — through the ordinary physics of building on a foundation that keeps moving underneath sixty thousand things that were each nailed down separately.

You can go and look at the result. Pull up a dozen WordPress sites at random. Slow, bloated, half of them with a layout breaking somewhere below the fold, and the ones that look good look good because somebody spent real money fighting the platform to get there. That isn't an insult to the idea. That is what twenty years of accumulated tolerance error looks like when you finally try to lift it.

04. Who Actually Got Hurt: The People Who Made the Bricks

I want to be careful here, because there is a version of this argument that curdles into contempt for the plugin authors, and they are the last people who deserve it.

They were handed a genuinely good pitch: build something useful, give it away, the ecosystem will look after you. Reputation, clients, maybe a pro tier one day. It sounded like a great idea. At the time it was a great idea.

What they actually signed up for was an unpaid maintenance treadmill on a platform whose release schedule they don't control. Every core update is a bill they never agreed to pay, arriving forever, in their own time. Thousands of people performing structural repair on somebody else's castle, indefinitely, in exchange for gratitude.

So when a plugin goes stale, or gets abandoned, or gets sold to a buyer who quietly turns it into an ad injector — those are not moral failures. That is the predictable end state of the deal they were offered. Burnout is not a character flaw when the arrangement had no exit in it.

05. The One Tier With Intent: Nobody Sells Glue for Real Legos

Which brings us to the tier where I am not feeling generous.

Zapier. Make. The whole integration layer. Billion-dollar businesses whose entire product is gluing together things that were supposed to work together already.

Understand what that means. That industry does not sell you a brick. It sells you the adapter between two bricks — a toll booth on a bridge that only needs to exist because the road was laid wrong. And it isn't some cottage trade at the edge of the ecosystem. It owns that ground. It is one of the most valuable positions in software, and the position is the gap.

Lego has no Zapier. There is no aftermarket in Lego adapters, no subscription that lets a 2×4 talk to a 2×2. Nothing that fits needs a paid translator. A thriving glue industry is not a feature of your stack. It is the receipt for its defect.

And then there is the price.

Ask the plain question: what am I buying? Something watches for an event and forwards some data to another service. That's the product. There is no human in the loop — not per task, not per month. Nobody at that company does anything when your invoice fires. The work is automated and unattended, and the automating of it is itself being automated. The marginal cost of serving you is approaching zero and still falling.

Meanwhile you are billed per task. You already pay for the first app. You already pay for the second app. Now a third company meters you every time the two of them speak.

A price that far untethered from cost has three explanations and only three: brand, scarcity, or a closing window. Nobody feels loyalty to a glue layer, and there is no shortage of HTTP requests.

06. The Standing Charge: You Never Owned Any of It

Count them sometime. Not the big ones — all of them. The form tool, the scheduler, the email thing, the popup thing, the CRM, the automation layer, the analytics add-on, the backup service, the security plugin's pro tier.

Thirteen thousand a year, forever, going up annually, and on the day you stop paying you have nothing. No asset, no fallback, no archive you can run. You rented every brick in the building and the rent never converts to equity.

And the bricks change shape while you sleep. Pricing changes. APIs deprecate. Free tiers close. A tool you built three years of process around is acquired and sunset on ninety days' notice. Occasionally a brick is simply recalled — the plugin is pulled, the maintainer walks, the foundation course of your castle disappears and you find out from a customer.

You are the systems integrator for a stack of twenty vendors. You were never hired for the job, never paid for it, and can't quit it.

07. In Defense of Everyone: Why We All Did It Anyway

Because for twenty years it was the right call. Genuinely.

Custom software meant eighty thousand dollars and nine months, minimum, and at the end you owned something that needed a developer on retainer to keep breathing. For a small business that wasn't a hard decision. It wasn't a decision at all. Cheap bricks that mostly fit beat no castle.

The duct tape was never stupidity. It was arithmetic. The numbers said: accept the tolerance error, it's the only thing you can afford.

08. The Premise Expired: The Arithmetic Changed

That's the part nobody has fully absorbed yet.

The entire justification for the knockoff pile was the crushing cost of the alternative. That cost collapsed. What used to be eighty thousand dollars and three quarters of a year is now weeks, at a fraction, and at the end of it the thing is yours. Your data, your schema, your logic, on your box. One authority over one spec. No rent. No vendor who gets to change the deal on a Tuesday.

09. Scope of the Collapse: What Actually Dies

Not all of it. Be precise here, because the sloppy version of this prediction is how you get dismissed.

Real infrastructure isn't going anywhere. Some software is genuinely hard, genuinely deep, and worth paying for on an ongoing basis. Pay for it.

What dies is the middle. The ninety-a-month tier for what is, structurally, a form and a database. The tool whose entire moat was that building it yourself used to be harder than paying for it. That moat is gone — not crossed, evaporated — and every business standing behind it is now priced against a thing anyone can simply have.

Software as a service is about fifteen years old. It is not the natural order of anything. Nobody rented their software in 2005. A business model that new isn't physics. It's a trend that ran long enough for a generation to mistake it for weather.

You are allowed to say out loud that you never agreed with it.