Your 7 business models may end up being the exact right ones for the current moment in time. But I can't help but feel that there is a potential flaw in the logic between your points 1 ("Software is getting cheaper and easier to build, and that will only continue.") and 2 ("Because of that, technical differentiation will be harder to create and maintain.").
The arc of time over which software has gotten cheaper and easier to build is very long, and way predates AI. Modern IDEs, cloud computing & hyperscaler building blocks, democratization of access to coding knowledge, better collaboration technology over the internet, Stack Overflow and similar, larger populations of eager developers in developing nations, etc.... these have lowered costs and barriers to building software and have been trends going over decades. (you know all this well, obviously)
But those reduced barriers have not led to decreased technical differentiation as far as I can tell. Rather it seems like we have sets of dominant software franchises just like we did in the 80s and 90s and 00s, only software is now much better, much more sophisticated and scalable and high-performance, and way more intricate. So rather than "cheaper/easier" implying "differentiation down, moats are lessened", it seems to have implied "product quality improves in lockstep with productivity improvements; moats and differentiation remain approx. unchanged in aggregate".
I think it would be a lot different if software development productivity improvements were uniquely accessible to new entrants in a way they weren't to incumbents (that would make it more like the cloud transition*, where it made more sense to start a new SaaS company than to retrofit an on-prem company to cloud). But as far as I can tell, the kinds of software development advances we're seeing in AI are equally available to incumbents (at least the ones who have adaptable cultures) and to entrants, which means that while, yes, new startups can build 0 to 1 products faster than ever before, incumbents can use AI to build further differentiation into their platforms faster and extend their leads. So that again would be the pattern of the quality bar rising while the ability to maintain differentiation is somewhat static.
I'd certainly be curious for your pushback here.
*(if it turns out that brand new AI-first software is just plain better at solving customers' problems than incumbent software with AI incorporated into existing workflows, then I agree that the differentiation dynamic would quickly shift away from existing franchises and make it harder for everyone to maintain share. But I don't think we know enough to say whether that's true or not.)
AI doesn’t make software “free” when scaling to thousands of users. That’s hardcore engineering, infra, security, uptime guarantees.
Scaling a real app isn’t a fairy tale, it’s architecture, infra, testing, cost control, and sweat. Your vision fits a 3‑button webapp, not real‑world scale.
This is probably the best piece I’ve read in the last couple of years on where software is heading.
What struck me most is how closely it aligns with the path we’re on at TACTUN—bundling software with hardware, eliminating R&D costs for our customers, and creating the kind of blue-ocean markets that didn’t exist before.
You’ve articulated the shift I’ve been sensing: the real moat is moving away from “just software” toward hardware-enabled platforms, convenience, and switching costs. Reading this felt like someone had put into words the strategy we’ve been living every day.
Really enjoyed this, Eric! Especially the framing of "software is dead, long live software." Other readers touched on this, but what we’ve seen in many medium-large sized organizations is that while AI has made generating code far easier, the real challenge at the enterprise level is everything around the code; architecture, DevOps, containers, UX, scaling plans, and long-term maintainability. We often meet teams after projects have stalled or because code alone doesn’t equal a resilient product. Your point about differentiation shifting away from pure software really resonated; that’s exactly where we see enterprises struggling to bridge the gap. Excited to keep following your writing on this.
As the cost of software components drops toward zero, I often compare it to the dropshipping model.
I think there’s a good chance that in the future, every KOL or small business owner will have their own app—whether it’s an AI agent, a mini-store, an ordering page, or a product site.
When everyone can sell, the products themselves matter less—the key is distribution.
Behind the scenes though, the big models, factories, and cloud providers will stay unified.
Front end = decentralized,
back end = centralized.
Software was never just service. At its core, it’s an expression.
This is just brilliant. I think you nailed it for the practice of law.
Your 7 business models may end up being the exact right ones for the current moment in time. But I can't help but feel that there is a potential flaw in the logic between your points 1 ("Software is getting cheaper and easier to build, and that will only continue.") and 2 ("Because of that, technical differentiation will be harder to create and maintain.").
The arc of time over which software has gotten cheaper and easier to build is very long, and way predates AI. Modern IDEs, cloud computing & hyperscaler building blocks, democratization of access to coding knowledge, better collaboration technology over the internet, Stack Overflow and similar, larger populations of eager developers in developing nations, etc.... these have lowered costs and barriers to building software and have been trends going over decades. (you know all this well, obviously)
But those reduced barriers have not led to decreased technical differentiation as far as I can tell. Rather it seems like we have sets of dominant software franchises just like we did in the 80s and 90s and 00s, only software is now much better, much more sophisticated and scalable and high-performance, and way more intricate. So rather than "cheaper/easier" implying "differentiation down, moats are lessened", it seems to have implied "product quality improves in lockstep with productivity improvements; moats and differentiation remain approx. unchanged in aggregate".
I think it would be a lot different if software development productivity improvements were uniquely accessible to new entrants in a way they weren't to incumbents (that would make it more like the cloud transition*, where it made more sense to start a new SaaS company than to retrofit an on-prem company to cloud). But as far as I can tell, the kinds of software development advances we're seeing in AI are equally available to incumbents (at least the ones who have adaptable cultures) and to entrants, which means that while, yes, new startups can build 0 to 1 products faster than ever before, incumbents can use AI to build further differentiation into their platforms faster and extend their leads. So that again would be the pattern of the quality bar rising while the ability to maintain differentiation is somewhat static.
I'd certainly be curious for your pushback here.
*(if it turns out that brand new AI-first software is just plain better at solving customers' problems than incumbent software with AI incorporated into existing workflows, then I agree that the differentiation dynamic would quickly shift away from existing franchises and make it harder for everyone to maintain share. But I don't think we know enough to say whether that's true or not.)
AI doesn’t make software “free” when scaling to thousands of users. That’s hardcore engineering, infra, security, uptime guarantees.
Scaling a real app isn’t a fairy tale, it’s architecture, infra, testing, cost control, and sweat. Your vision fits a 3‑button webapp, not real‑world scale.
Love this article - have Restacked it for my audience ✨
This is probably the best piece I’ve read in the last couple of years on where software is heading.
What struck me most is how closely it aligns with the path we’re on at TACTUN—bundling software with hardware, eliminating R&D costs for our customers, and creating the kind of blue-ocean markets that didn’t exist before.
You’ve articulated the shift I’ve been sensing: the real moat is moving away from “just software” toward hardware-enabled platforms, convenience, and switching costs. Reading this felt like someone had put into words the strategy we’ve been living every day.
Really enjoyed this, Eric! Especially the framing of "software is dead, long live software." Other readers touched on this, but what we’ve seen in many medium-large sized organizations is that while AI has made generating code far easier, the real challenge at the enterprise level is everything around the code; architecture, DevOps, containers, UX, scaling plans, and long-term maintainability. We often meet teams after projects have stalled or because code alone doesn’t equal a resilient product. Your point about differentiation shifting away from pure software really resonated; that’s exactly where we see enterprises struggling to bridge the gap. Excited to keep following your writing on this.
As the cost of software components drops toward zero, I often compare it to the dropshipping model.
I think there’s a good chance that in the future, every KOL or small business owner will have their own app—whether it’s an AI agent, a mini-store, an ordering page, or a product site.
When everyone can sell, the products themselves matter less—the key is distribution.
Behind the scenes though, the big models, factories, and cloud providers will stay unified.
Front end = decentralized,
back end = centralized.
Software was never just service. At its core, it’s an expression.