We might see the rise of the Product Engineer (PE). These PEs will form highly talented, lean teams where members are largely interchangeable.
Rather than splitting work along technical expertise (like frontend, backend, or infrastructure) teams will organize around products and features. In this model, engineers aren’t siloed by stack; they’re aligned by outcomes. One PE might work on onboarding, another payments, another notifications; each accountable for their feature end-to-end... The structure shifts from "frontend/backend/infra teams" to "feature squads", each with full-stack autonomy.
Andrew, this is such a thought-provoking post. Your "Home Screen Test" is a perfect way to show how early we still are in this space. The questions you've laid out are the exact ones everyone in startups is wrestling with right now. I completely agree that the next few years will be about building the business logic and UIs on top of foundation models, not the models themselves.
On that note, it's cool to see companies like Nutron.ai emerge. They're focused on building those custom AI solutions and workflows for businesses, which is exactly what your post is talking about. It's like they're enabling companies to be those "super-productive" teams that you're hoping for in the most optimistic view.
I think the home screen test might be missing the bigger picture.
Instead of counting AI-native apps, we should be looking at:
→ Time spent: Are we using those 28 apps less because LLM apps handle multiple use cases?
→ Usage reduction: How many tasks are now consolidated into fewer interfaces?
My prediction: We're heading toward one super-app that handles travel, food, commerce through MCP connectors and chat interfaces. Invisible design where you don't need 28 different UIs.
The real AI transformation is happening backstage first:
• Orchestra layer optimisation
• Backend service intelligence
• Frontend will be the last to pivot
Counting app icons feels incremental when the entire interaction paradigm is shifting toward conversational interfaces.
What we're building isn't 28 AI apps - it's the infrastructure for a world where we do not need so many apps.
People will be laid off. 4x overstaffed companies will become 2x overstaffed companies. But consumer demand for apps will not change. The capacity of marketing new digital products will not change. The rate at which big businesses adopt new B2B products will not change. The VC funding stuff will not change. All of these will become bottlenecks, and the incumbents will be best positioned to push their stuff through these bottlenecks first.
The IT job market might crash and millions of junior software developers will need to find a real job. That's it.
Its true using AI already make your work 3x faster. Suggestion for any startup; provide solutions to other people's problems. People pay millions to solve problems that are so simple they don't even realize how easy they are, and what's better than using AI to find many practical solutions.
"One easy answer is that the last decade of consumer apps has taught us that, in a low technical differentiation ecosystem, user growth and network effects are all that matter."
"The argument for continued decentralization is that if building products becomes trivially easy, it's almost like a form of content creation. And as we've seen with content creators, there are local and global creators who can be based anywhere."
"The next few years will be about the folks who build the business logic that sits on top of these models. They won't do AI research or train their own foundation models."
This is such a great post because it really asks the right questions - thanks Andrew. There is another lesson of economic history. New technologies and greater wealth create more specialisation of labour. So AI will likely collapse some team distinctions eg product and engineering. Those roles as we know them today will disappear (with much wailing from the doomsters) but they will be replaced by new jobs with new, finer grained specialisms. Impossible to know what those will be but the economy will adapt and grow as a result.
The real shift will come when AI isn’t a feature but the foundation: an AI native calendar that proactively manages your time, or a social network that dynamically adapts to your relationships. We’re still waiting for that paradigm flip, where AI enables not just better tools, but entirely new behaviors. The infrastructure is here; the imagination is catching up.
Its interesting that you seem to be betting more on killer features than distribution. Wouldn't an average feature by Meta or even LinkedIn/Snap more likely to reach scale than by a small start up in Sweden or Indonesia or Kenya? Or are you saying that there need to be a fragmented growth before consolidation happens again?
Startups exploit cracks in incumbent business models.
Innovation waves historically moved along single axes:
- 1960s-80s: Databases
- 1970s-90s: Computing power
- 1990s-2010s: Internet connectivity
- 2010s-20s: Blockchain/decentralization
Today's difference: Multiple axes advance simultaneously (AI, quantum, biotech, AR/VR, IoT) while the target—value creation—stays fixed. It's like shooting arrows while spinning on an accelerating platform.
Winning strategy: Master one axis. Form strong opinions about the others. Build assuming they'll converge with yours.
AI will help us get further faster if we already know what we want and how to build it. Instead of less developers and more code. We should think of all the things we can do now that we would not try to do given a teams capabilities in the past. We should be hiring not firing and making more software solutions.
As a startup, I build an app that uses AI for decision making in code and used for creating some of the code. I have AI do my repetitive typing tasks for me. One developer was able to build this from scratch in 4 months. https://www.recipe2kitchen.com
> How will we organize the team of the future?
We might see the rise of the Product Engineer (PE). These PEs will form highly talented, lean teams where members are largely interchangeable.
Rather than splitting work along technical expertise (like frontend, backend, or infrastructure) teams will organize around products and features. In this model, engineers aren’t siloed by stack; they’re aligned by outcomes. One PE might work on onboarding, another payments, another notifications; each accountable for their feature end-to-end... The structure shifts from "frontend/backend/infra teams" to "feature squads", each with full-stack autonomy.
I was thinking about this idea a while back and compiled some thoughts here: https://nandinfinitum.com/posts/the-product-engineer/
Andrew, this is such a thought-provoking post. Your "Home Screen Test" is a perfect way to show how early we still are in this space. The questions you've laid out are the exact ones everyone in startups is wrestling with right now. I completely agree that the next few years will be about building the business logic and UIs on top of foundation models, not the models themselves.
On that note, it's cool to see companies like Nutron.ai emerge. They're focused on building those custom AI solutions and workflows for businesses, which is exactly what your post is talking about. It's like they're enabling companies to be those "super-productive" teams that you're hoping for in the most optimistic view.
I think the home screen test might be missing the bigger picture.
Instead of counting AI-native apps, we should be looking at:
→ Time spent: Are we using those 28 apps less because LLM apps handle multiple use cases?
→ Usage reduction: How many tasks are now consolidated into fewer interfaces?
My prediction: We're heading toward one super-app that handles travel, food, commerce through MCP connectors and chat interfaces. Invisible design where you don't need 28 different UIs.
The real AI transformation is happening backstage first:
• Orchestra layer optimisation
• Backend service intelligence
• Frontend will be the last to pivot
Counting app icons feels incremental when the entire interaction paradigm is shifting toward conversational interfaces.
What we're building isn't 28 AI apps - it's the infrastructure for a world where we do not need so many apps.
My prediction: AI will not change much.
People will be laid off. 4x overstaffed companies will become 2x overstaffed companies. But consumer demand for apps will not change. The capacity of marketing new digital products will not change. The rate at which big businesses adopt new B2B products will not change. The VC funding stuff will not change. All of these will become bottlenecks, and the incumbents will be best positioned to push their stuff through these bottlenecks first.
The IT job market might crash and millions of junior software developers will need to find a real job. That's it.
Its true using AI already make your work 3x faster. Suggestion for any startup; provide solutions to other people's problems. People pay millions to solve problems that are so simple they don't even realize how easy they are, and what's better than using AI to find many practical solutions.
This is an insightful estimation.
My favorite parts are:
"One easy answer is that the last decade of consumer apps has taught us that, in a low technical differentiation ecosystem, user growth and network effects are all that matter."
"The argument for continued decentralization is that if building products becomes trivially easy, it's almost like a form of content creation. And as we've seen with content creators, there are local and global creators who can be based anywhere."
"The next few years will be about the folks who build the business logic that sits on top of these models. They won't do AI research or train their own foundation models."
Also, here's my learning: https://glasp.co/kei/p/94afcfedc04765de9536
by far my fav read this week, thanks for your wisdom as always!!
All very good questions!
This is such a great post because it really asks the right questions - thanks Andrew. There is another lesson of economic history. New technologies and greater wealth create more specialisation of labour. So AI will likely collapse some team distinctions eg product and engineering. Those roles as we know them today will disappear (with much wailing from the doomsters) but they will be replaced by new jobs with new, finer grained specialisms. Impossible to know what those will be but the economy will adapt and grow as a result.
The real shift will come when AI isn’t a feature but the foundation: an AI native calendar that proactively manages your time, or a social network that dynamically adapts to your relationships. We’re still waiting for that paradigm flip, where AI enables not just better tools, but entirely new behaviors. The infrastructure is here; the imagination is catching up.
Its interesting that you seem to be betting more on killer features than distribution. Wouldn't an average feature by Meta or even LinkedIn/Snap more likely to reach scale than by a small start up in Sweden or Indonesia or Kenya? Or are you saying that there need to be a fragmented growth before consolidation happens again?
Startups exploit cracks in incumbent business models.
Innovation waves historically moved along single axes:
- 1960s-80s: Databases
- 1970s-90s: Computing power
- 1990s-2010s: Internet connectivity
- 2010s-20s: Blockchain/decentralization
Today's difference: Multiple axes advance simultaneously (AI, quantum, biotech, AR/VR, IoT) while the target—value creation—stays fixed. It's like shooting arrows while spinning on an accelerating platform.
Winning strategy: Master one axis. Form strong opinions about the others. Build assuming they'll converge with yours.
The magic happens at intersections.
AI will help us get further faster if we already know what we want and how to build it. Instead of less developers and more code. We should think of all the things we can do now that we would not try to do given a teams capabilities in the past. We should be hiring not firing and making more software solutions.
As a startup, I build an app that uses AI for decision making in code and used for creating some of the code. I have AI do my repetitive typing tasks for me. One developer was able to build this from scratch in 4 months. https://www.recipe2kitchen.com
Hi andrew, the end of the first bulley point is missing. Thank you for your posts !