In one week, three voices who don’t know each other posted the same diagnosis from different microphones. Farza tweeted on Tuesday with 502K views: anyone can bake bread at home, yet 99% of us still buy it. Eduardo Ordax, Generative AI Lead at AWS with 200K followers, restated the analogy on Thursday and added 1,660 reactions on top. Jaya Gupta of Foundation Capital published a piece on Friday that crossed two million views in 24 hours and went one floor deeper.

The first two reassure: software is not dead. The third is the one that decides anything. When everything copyable becomes cheap, the only thing left is the shape of the institution producing the work. Last week’s distribution piece looked at the floor below: how to deliver what you’ve already built. This one looks at the floor above: what kind of company can produce it.

Why the bread analogy falls short

In the Ordax thread, the comments worth reading are the objections. Bo Bleyl, a software architect, put it in one line: replacing a SaaS subscription with AI is usually a half-day, one-off effort, and it stays good indefinitely. Bread is baked again tomorrow; AI-built software is baked once. The analogy breaks at frequency.

Cameron Palmer, an AI/LLM integration engineer, sharpened it further. The hard part isn’t baking the loaf; it’s knowing which loaves are bad before they reach the customer. That’s where most production AI efforts actually stall. Not at the building, at the evaluation. Evaluation is a shape function. It’s the kill rate, the decision rights over what gets killed and what gets scaled, the named owner of “good enough to ship.” None of that gets bought.

Being chosen vs being seen

The sharpest line in Jaya’s essay, the one worth reading twice, is structural: “you are paying in identity what you do not want to pay in structure.”

Ambitious people, she argues, want to feel six different things: special, destined for something larger, in the room where FOMO can’t catch them, with something to prove, close to power, capable of the kind of sacrifice that makes work feel like meaning. Each of those six is an emotional promise. And every emotional promise, to actually be true, has to be backed by a structural one: scope, authority, economic participation, decision rights.

If a company says customer proximity is the moat but customer-facing work is low-status internally, the promise is fake. The same is true when it says ownership matters and centralizes every meaningful decision in the executive committee. Worst of all is a mission that offends no one, selects for no one, and costs no one anything.

From there comes the distinction. Being chosen is emotional: she’s told that she’s special, that the team trusts her, that she belongs. Being seen is structural: her scope is named, her authority is in writing, her economic participation has a triggering event, her decision rights show up in some document the team refers back to when there’s a disagreement. A company that only delivers the first captures founder-level intensity and pays back in belonging. When the structure never catches up, the promise gets denominated in time. Time never announces when it leaves.

Five cycles since 1990. What survives is the shape.

I’ve been in computing for 36 years. I started at 15 on a Commodore 64. I’ve watched this pattern five times: a layer becomes cheap, differentiation moves up one floor, and the company that survives the cycle is not the one with the best product of the moment. It’s the one whose shape the competitor can’t copy even after they’ve copied the product the same quarter.

Cheap hardware in the early 90s: the learning-shaped company won. Cheap internet distribution in the late 90s: the brand with pull won. Cheap leads from SEO and SEM in the mid-2000s: the product with retention won. Cheap reach from organic social through 2020, until iOS 14.5 broke Meta’s ROAS overnight in April 2021: the company with its own audience won. In 2026, building is what’s cheap. Differentiation moves to who decides what to build, how the team is organized, how output gets evaluated. The institution-shaped company survives. The roadmap-shaped one does not. I covered the upstream face and the external face; this is the internal one.

Five questions for Monday’s leadership meeting

1. Who has customer proximity AND status simultaneously? If your strongest engineer never talks to a customer and your CSM has no authority over the roadmap, you don’t have a customer-oriented company. You have two companies stapled together.

2. Where do decision rights actually live, at the edge or in the all-hands? If the last meaningful technical decision waited three weeks for a meeting where half the room hadn’t read the code, your iteration cycle runs at the speed of meetings. Not at the speed of your best people.

3. What kind of person can only become themselves at this company? If you can’t name them in a single specific sentence (“the engineer with three years in product and two in operations who wants to build the system both teams always needed”), you don’t have a shape. You have a job board.

4. Which promise to your best people is denominated in time without a concrete mechanism behind it? Phrases like “you’ll grow with the company” need a mechanism: a quarterly review with written criteria, a scope that expands when a named trigger fires, or economic participation tied to a specific event. Without a mechanism, the promise is pure identity. And pure identity walks out the door the day a competitor offers the same thing, backed by structure.

5. When AI compresses your build layer 5x next quarter, which of the four answers above changes? If all four hold, your shape is independent of the product. That’s the test. If any of them depends on having fifteen engineers doing what three could do tomorrow with Claude, that part was never shape. It was an artifact of building being expensive.

At IQ Source we run AI Maestro before the next quarter goes into expanding model context or hiring another vendor. The audit doesn’t ask, “does your team use Claude?” It asks whether your company’s shape produces the judgment, ownership and velocity that AI demands. For software companies whose product is part of the customer’s surface, Tech Partner is the structural answer to the bread analogy: you partner not because you can’t bake, but because the shape you’d need to bake everything yourself isn’t the shape that wins your category. The commoditized layer (build, infra, runtime; covered in April) IQ Source runs. The non-commoditized layer stays in your house.

What kind of company can only be yours? Which promise you made to your best people this week is denominated in identity rather than in structure? The team that answers both pays one in 2027. The team that doesn’t ends up paying three, and explaining to its investor why a competitor with a smaller product walked off with the customer.