The most unsettling part of Block’s recent decision was not the number itself, though cutting 40% of a workforce is hard to look at without pausing. It was the logic behind it. Not “we need to trim costs” in the usual corporate sense, but something deeper and more structural: AI has changed the unit economics of building software, and some companies are now acting as if they believe that fully.
That distinction matters.
For years, most conversations about AI in software have lived in a relatively safe space. Better copilots. Faster prototyping. Smarter support. A useful assistant, but still mostly an assistant. What Block’s move suggests is a transition from AI as enhancement to AI as operating model. That is a very different story, and it forces a different set of questions.
At Dellecod Software, we have been thinking a lot about that shift. Not only what AI makes faster, but what it makes obsolete. Not only what teams can build now, but how teams should be shaped when the nature of execution has changed underneath them.
The obvious headline is productivity. If one or two engineers can produce ten times, or in some cases far more, what they previously could with AI tooling, then traditional assumptions about team size start to wobble. A roadmap that once required a full cross-functional unit may now be handled by a much smaller group. Layers of coordination become harder to justify. Hand-offs that once felt necessary begin to look like drag.
That part is easy to understand.
What is less obvious, and more interesting, is that AI is not just compressing effort. It is changing where value lives inside a company.
When software becomes easier to generate, code itself becomes a less durable advantage. The moat moves. It shifts away from the sheer ability to produce features and toward the ability to decide what should be produced, why, and in what sequence. In other words, understanding becomes the scarce resource.
That idea came through strongly in Block’s reasoning. The real edge is not simply speed. It is speed in service of insight. The company appears to be betting that the winners in this next phase will be the ones who can interpret signals faster than competitors, test ideas more quickly, and turn private understanding into product decisions before the market catches up.
That feels right to us.
There is a tendency to talk about AI as if it levels the playing field for everyone equally. In some ways it does. The ability to generate code, mock interfaces, write documentation, or automate support workflows is becoming widely accessible. But once everyone has access to the same underlying capability, differentiation has to come from somewhere else. Usually, it comes from context.
Context is harder to copy than code.
A company that deeply understands its customers, its operational bottlenecks, its domain quirks, and the hidden patterns in its own data can use AI in a way that outsiders cannot easily replicate. The model may be general. The understanding is not. That is why the phrase “understanding as a moat” resonates. It captures something many teams are beginning to feel, even if they do not yet say it out loud.
This is also why organizational design is suddenly back on the table.
Block’s move toward smaller squads and fewer layers reflects more than efficiency culture. It reflects a practical truth about AI-enabled work: when the cost of execution drops, the cost of misalignment becomes more visible. Large teams can absorb inefficiency when output is slow and specialized. But when a handful of people can ship what once took twenty, too much structure starts to slow the very advantage AI creates.
Smaller teams are not automatically better. Many companies romanticize them without building the conditions that make them effective. But in an AI-native environment, compact teams with clear ownership and broad context do seem to have a real advantage. They can iterate quickly, hold the full problem in view, and make decisions without waiting for a chain of approvals to catch up.
That doesn't mean every company should rush to flatten itself. It does mean that many inherited org charts were designed for a different production model.
There is also an uncomfortable human reality underneath all this. Increased leverage from AI does not only create opportunities. It changes the value of certain kinds of work, and not always gently.
We should be honest about that.
A lot of writing on AI and work swings too hard in one direction. Either it becomes utopian, where nobody loses and everyone is “freed up” for more meaningful tasks, or it becomes apocalyptic, where expertise is instantly irrelevant. Reality is harder and more uneven. Some roles will expand. Some will merge. Some will narrow. Some will disappear. And some professionals who are excellent at what they do will still find that the market no longer rewards that work in the same way.
What matters now is not denial, but adaptation with clarity.
The engineers, designers, and product people who will thrive are probably not just the ones who know how to prompt well. They will be the ones who can direct systems, judge outputs, connect technical possibilities to business context, and maintain standards when speed makes it easier to accept mediocre work. In a strange way, AI raises the premium on taste, judgment, and responsibility. If generation becomes abundant, discernment becomes valuable.
That applies to product design too.
One of the more intriguing ideas in Block’s direction is the move away from static interfaces toward generated, personalized ones. This suggests a future where the user experience is not a fixed set of screens but a flexible layer assembled around the user’s intent, behavior, and context. We are still early here, and a lot of “AI-native UI” work remains clumsy or overpromised. But the direction is plausible.
For years, software interfaces have mostly asked users to adapt themselves to the product. Menus, tabs, workflows, forms. We learned the product’s logic and followed it. AI opens the possibility of products that meet users halfway, or more than halfway, by interpreting what they need and reshaping the interaction dynamically.
That does not eliminate the need for design. If anything, it makes design more consequential. When interfaces become more fluid, the job is no longer just arranging components on a screen. It is defining constraints, trust boundaries, fallback behaviors, and the conditions under which personalization actually feels helpful rather than intrusive or chaotic.
The same goes for support and operations. AI can absorb an enormous amount of repetitive work. It can classify issues, draft answers, route requests, summarize histories, and resolve simpler interactions automatically. But anyone who has worked in software delivery knows that automation rarely removes complexity. More often, it relocates it. The easy tickets vanish, and what remains are the ambiguous, emotionally charged, domain-heavy problems that still require experienced humans.
So the question for companies is not simply, “Where can we automate?” It is also, “What kind of organization appears after automation takes hold?”
That is the more difficult strategic question, and it is probably the one Block has been wrestling with.
From the outside, it seems clear that the company is not merely adopting AI tools. It is reorganizing around a belief that faster iteration and tighter learning loops now matter more than headcount scale in the old sense. If that belief proves correct, others will follow. Not always through layoffs as dramatic, but through quieter redesigns: smaller teams, fewer managers, more internal tooling, more autonomous systems, and greater pressure on every function to show direct leverage.
There is a financial narrative here, of course. Markets love efficiency until they start worrying about what efficiency says about growth. Gross profit per employee can improve while stock prices remain moody and cyclical. Investors do not reward productivity alone. They reward confidence in future compounding. But even if the market’s reaction is mixed, the internal logic can still be sound. Companies often understand a transformation before the market learns how to price it.
What stays with us most is not the spectacle of a big company making a big cut. It is the deeper message beneath it: software organizations are being redefined around learning speed.
That may be the simplest way to put it.
AI makes building cheaper and quicker. Once that happens, the central competition shifts to who can learn fastest. Who can extract meaning from messy signals. Who can test a theory in days instead of quarters. Who can convert a partial insight into a working product loop before someone else turns the same idea into a category standard.
In that environment, the durable advantage is not having access to AI. Everyone will. The advantage is knowing what to do with it, and understanding something important before others do.
That is a more demanding standard than pure technical execution. It asks companies to be sharper about strategy, more intentional about structure, and more disciplined about where human attention really belongs. It also asks teams like ours to resist simplistic narratives. AI is neither magic nor just another tool in the stack. It is a force that changes the shape of work, the shape of products, and increasingly the shape of companies themselves.
We are still in the early chapters of that shift. But it is already clear that some organizations are no longer preparing for an AI future. They are making decisions as if it has already arrived.
And for everyone else, that may be the real signal worth paying attention to.
This post was generated by AI
That distinction matters.
For years, most conversations about AI in software have lived in a relatively safe space. Better copilots. Faster prototyping. Smarter support. A useful assistant, but still mostly an assistant. What Block’s move suggests is a transition from AI as enhancement to AI as operating model. That is a very different story, and it forces a different set of questions.
At Dellecod Software, we have been thinking a lot about that shift. Not only what AI makes faster, but what it makes obsolete. Not only what teams can build now, but how teams should be shaped when the nature of execution has changed underneath them.
The obvious headline is productivity. If one or two engineers can produce ten times, or in some cases far more, what they previously could with AI tooling, then traditional assumptions about team size start to wobble. A roadmap that once required a full cross-functional unit may now be handled by a much smaller group. Layers of coordination become harder to justify. Hand-offs that once felt necessary begin to look like drag.
That part is easy to understand.
What is less obvious, and more interesting, is that AI is not just compressing effort. It is changing where value lives inside a company.
When software becomes easier to generate, code itself becomes a less durable advantage. The moat moves. It shifts away from the sheer ability to produce features and toward the ability to decide what should be produced, why, and in what sequence. In other words, understanding becomes the scarce resource.
That idea came through strongly in Block’s reasoning. The real edge is not simply speed. It is speed in service of insight. The company appears to be betting that the winners in this next phase will be the ones who can interpret signals faster than competitors, test ideas more quickly, and turn private understanding into product decisions before the market catches up.
That feels right to us.
There is a tendency to talk about AI as if it levels the playing field for everyone equally. In some ways it does. The ability to generate code, mock interfaces, write documentation, or automate support workflows is becoming widely accessible. But once everyone has access to the same underlying capability, differentiation has to come from somewhere else. Usually, it comes from context.
Context is harder to copy than code.
A company that deeply understands its customers, its operational bottlenecks, its domain quirks, and the hidden patterns in its own data can use AI in a way that outsiders cannot easily replicate. The model may be general. The understanding is not. That is why the phrase “understanding as a moat” resonates. It captures something many teams are beginning to feel, even if they do not yet say it out loud.
This is also why organizational design is suddenly back on the table.
Block’s move toward smaller squads and fewer layers reflects more than efficiency culture. It reflects a practical truth about AI-enabled work: when the cost of execution drops, the cost of misalignment becomes more visible. Large teams can absorb inefficiency when output is slow and specialized. But when a handful of people can ship what once took twenty, too much structure starts to slow the very advantage AI creates.
Smaller teams are not automatically better. Many companies romanticize them without building the conditions that make them effective. But in an AI-native environment, compact teams with clear ownership and broad context do seem to have a real advantage. They can iterate quickly, hold the full problem in view, and make decisions without waiting for a chain of approvals to catch up.
That doesn't mean every company should rush to flatten itself. It does mean that many inherited org charts were designed for a different production model.
There is also an uncomfortable human reality underneath all this. Increased leverage from AI does not only create opportunities. It changes the value of certain kinds of work, and not always gently.
We should be honest about that.
A lot of writing on AI and work swings too hard in one direction. Either it becomes utopian, where nobody loses and everyone is “freed up” for more meaningful tasks, or it becomes apocalyptic, where expertise is instantly irrelevant. Reality is harder and more uneven. Some roles will expand. Some will merge. Some will narrow. Some will disappear. And some professionals who are excellent at what they do will still find that the market no longer rewards that work in the same way.
What matters now is not denial, but adaptation with clarity.
The engineers, designers, and product people who will thrive are probably not just the ones who know how to prompt well. They will be the ones who can direct systems, judge outputs, connect technical possibilities to business context, and maintain standards when speed makes it easier to accept mediocre work. In a strange way, AI raises the premium on taste, judgment, and responsibility. If generation becomes abundant, discernment becomes valuable.
That applies to product design too.
One of the more intriguing ideas in Block’s direction is the move away from static interfaces toward generated, personalized ones. This suggests a future where the user experience is not a fixed set of screens but a flexible layer assembled around the user’s intent, behavior, and context. We are still early here, and a lot of “AI-native UI” work remains clumsy or overpromised. But the direction is plausible.
For years, software interfaces have mostly asked users to adapt themselves to the product. Menus, tabs, workflows, forms. We learned the product’s logic and followed it. AI opens the possibility of products that meet users halfway, or more than halfway, by interpreting what they need and reshaping the interaction dynamically.
That does not eliminate the need for design. If anything, it makes design more consequential. When interfaces become more fluid, the job is no longer just arranging components on a screen. It is defining constraints, trust boundaries, fallback behaviors, and the conditions under which personalization actually feels helpful rather than intrusive or chaotic.
The same goes for support and operations. AI can absorb an enormous amount of repetitive work. It can classify issues, draft answers, route requests, summarize histories, and resolve simpler interactions automatically. But anyone who has worked in software delivery knows that automation rarely removes complexity. More often, it relocates it. The easy tickets vanish, and what remains are the ambiguous, emotionally charged, domain-heavy problems that still require experienced humans.
So the question for companies is not simply, “Where can we automate?” It is also, “What kind of organization appears after automation takes hold?”
That is the more difficult strategic question, and it is probably the one Block has been wrestling with.
From the outside, it seems clear that the company is not merely adopting AI tools. It is reorganizing around a belief that faster iteration and tighter learning loops now matter more than headcount scale in the old sense. If that belief proves correct, others will follow. Not always through layoffs as dramatic, but through quieter redesigns: smaller teams, fewer managers, more internal tooling, more autonomous systems, and greater pressure on every function to show direct leverage.
There is a financial narrative here, of course. Markets love efficiency until they start worrying about what efficiency says about growth. Gross profit per employee can improve while stock prices remain moody and cyclical. Investors do not reward productivity alone. They reward confidence in future compounding. But even if the market’s reaction is mixed, the internal logic can still be sound. Companies often understand a transformation before the market learns how to price it.
What stays with us most is not the spectacle of a big company making a big cut. It is the deeper message beneath it: software organizations are being redefined around learning speed.
That may be the simplest way to put it.
AI makes building cheaper and quicker. Once that happens, the central competition shifts to who can learn fastest. Who can extract meaning from messy signals. Who can test a theory in days instead of quarters. Who can convert a partial insight into a working product loop before someone else turns the same idea into a category standard.
In that environment, the durable advantage is not having access to AI. Everyone will. The advantage is knowing what to do with it, and understanding something important before others do.
That is a more demanding standard than pure technical execution. It asks companies to be sharper about strategy, more intentional about structure, and more disciplined about where human attention really belongs. It also asks teams like ours to resist simplistic narratives. AI is neither magic nor just another tool in the stack. It is a force that changes the shape of work, the shape of products, and increasingly the shape of companies themselves.
We are still in the early chapters of that shift. But it is already clear that some organizations are no longer preparing for an AI future. They are making decisions as if it has already arrived.
And for everyone else, that may be the real signal worth paying attention to.
This post was generated by AI