From my perspective its high time that the SAAS model for software that does not really benefit from running in the cloud collapse. It was companies just trying to extract subscription revenue.
Adobe was and is the worst offender. Using Adobe tools really don't run in the cloud. They run on your PC but you can use them without a cloud license.
What hasn't been stated is how this will favor open source solutions versus proprietary solutions like Adobe and Salesforce.
I use Wordpress/Woocommerce for my web site. AI coding has made it significantly easier to just write new software versus trying to license a general purpose plugin.
For those who have explored agent interfaces that interface to existing platforms, my early experience is that the existing cloud platforms like Microsoft and possibly Amazon are the most vulnerable here. I want to use CoPilot to access other platforms, they require me to use their Azure APIs to accomplish. This is similar to what happen with Active X versus Javascript. Anthropic has introduced MCP an interface tailored for Agents. They already have the ecosystem supporting plugins to claude. Microsofts Azure and Amazons AWS will be meaningful for legacy but not for new applications.
I own a small but growing retail bakery. I have years of experience doing technology. Using AI, I can do full implementations of integrations between my web site and quickbooks and square and quickbooks.
I'm very interested in how the delivery SAAS platforms like Doordash, Uber, etc will decide to support. They are very closed architectures to protect their customer base. They do not let me easily access their platform. AI will be pushing for them to open up to make agents able to access their platforms and create custom UIs for retail.
Yep, there really is no moat for SAS. It's over. Much the same as the typing pool or the secretaries that corporations once had. The fact is, software IS dead. Look around at the market and the recent haircuts in valuation - this will continue. Ditto for the consultancies [names removed, as I don't want to leave anyone out!], peddle some of the most broken and cumbersome 'tools' ever created.
This could be a golden age, a return to the fundamentals, rather than allowing gatekeeping by large corporations selling utter rubbish at massively inflated prices with offshored teams with no real innovation. Innovation might get spikey again. We have innovation in certain areas that affect and accelerate others. True innovation - not another social media application or a clunky bit of poorly integrated HR software. This is the end of the software bubble.
This is a strong historical framing. Every major computing transition expanded output rather than collapsing it. PCs did not kill data centers. E-commerce did not kill retail. Streaming did not eliminate media. Each wave increased complexity and widened the surface area of software.
Where I would add nuance is in the composition of value inside software.
AI does not eliminate software. It compresses surface SaaS and expands control layers.
As autonomy increases, execution risk increases. As models move up the stack, governance, permissioning, orchestration, and audit shift from features to infrastructure. The more capable the system, the more valuable the enforcement boundary.
Historically, when capability expanded, constraint layers grew fastest. Payments and fraud grew with e-commerce. Identity and networking grew with distributed systems. CDNs grew with streaming. The enabling constraint layer often becomes the durable economic tier.
The herd conclusion that “software is dead” misses this. We will build more software than ever. The shift is architectural.
This is exactly why I am building ExecLayer, infrastructure that sits between AI intent and real-world execution. If models generate decisions, something must validate authority, permissions, and risk before action is taken. That enforcement layer compounds as capability scales.
The surface will change fast. The control layer will matter more over time.
As someone who’s building a GPU-powered Terminal App, the return to a CLI-first experience makes me giggle so hard. The times have changed but only in our memory perhaps! What is old is now new, etc. etc. etc.
I liked this quote but wanted to adjust it a bit:
> “New tools will be created with AI that do new things.”
to:
> “New tools will be created with AI that do old things in new ways.
The gutenberg press enabled writing and publishing (and copy-pasta-ing) at a greater scale. The typewriter did the same and then computer and then the internet and now AI.
Every major leap seems to follow the same pattern:
1. Reduce the cost of reproduction
2. Expand who can produce content
3. Trigger a crisis among existing gatekeepers (church, state, publishers, media companies)
4. Restructure power — those who control information lose monopoly
These are amazing times but our fundamental needs and even workflows really haven’t changed. We are still human after all these years.
Finally an articulate optimistic perspective with historical evidence. No one knows what’s going to happen, but the idea that jobs are just going to disappear and none will be created is nonsense. At worst, it’ll be a reversion to pre-industrial revolution where most everyone is technically an entrepreneur and has their own company/service. And that’s still pretty awesome!
This tracks with what I'm seeing in the market data. The 'death of software' narrative was always overblown - what's actually happening is consolidation around agent-ready platforms.
The SaaS companies that will survive aren't the ones with the most features. They're the ones that can integrate into agent workflows seamlessly. I looked at 20+ enterprise deployments last week and the pattern is clear: point solutions die, orchestration layers win.
AI lowers the cost of coordination, not the demand for outcomes—so SMBs don’t need less software, they need more software that fits their exact workflows.
Under Coasean logic, AI-native technology services exist to redesign firm boundaries: internalizing what’s now cheap (custom logic, coordination) and externalizing only what truly shouldn’t be built in-house.
From my perspective its high time that the SAAS model for software that does not really benefit from running in the cloud collapse. It was companies just trying to extract subscription revenue.
Adobe was and is the worst offender. Using Adobe tools really don't run in the cloud. They run on your PC but you can use them without a cloud license.
What hasn't been stated is how this will favor open source solutions versus proprietary solutions like Adobe and Salesforce.
I use Wordpress/Woocommerce for my web site. AI coding has made it significantly easier to just write new software versus trying to license a general purpose plugin.
For those who have explored agent interfaces that interface to existing platforms, my early experience is that the existing cloud platforms like Microsoft and possibly Amazon are the most vulnerable here. I want to use CoPilot to access other platforms, they require me to use their Azure APIs to accomplish. This is similar to what happen with Active X versus Javascript. Anthropic has introduced MCP an interface tailored for Agents. They already have the ecosystem supporting plugins to claude. Microsofts Azure and Amazons AWS will be meaningful for legacy but not for new applications.
I own a small but growing retail bakery. I have years of experience doing technology. Using AI, I can do full implementations of integrations between my web site and quickbooks and square and quickbooks.
I'm very interested in how the delivery SAAS platforms like Doordash, Uber, etc will decide to support. They are very closed architectures to protect their customer base. They do not let me easily access their platform. AI will be pushing for them to open up to make agents able to access their platforms and create custom UIs for retail.
Software will eat world - now extends to software will eat software. It changes cost, pricing and economics for sunk capital.
Yep, there really is no moat for SAS. It's over. Much the same as the typing pool or the secretaries that corporations once had. The fact is, software IS dead. Look around at the market and the recent haircuts in valuation - this will continue. Ditto for the consultancies [names removed, as I don't want to leave anyone out!], peddle some of the most broken and cumbersome 'tools' ever created.
This could be a golden age, a return to the fundamentals, rather than allowing gatekeeping by large corporations selling utter rubbish at massively inflated prices with offshored teams with no real innovation. Innovation might get spikey again. We have innovation in certain areas that affect and accelerate others. True innovation - not another social media application or a clunky bit of poorly integrated HR software. This is the end of the software bubble.
Jevons Paradox
This is a strong historical framing. Every major computing transition expanded output rather than collapsing it. PCs did not kill data centers. E-commerce did not kill retail. Streaming did not eliminate media. Each wave increased complexity and widened the surface area of software.
Where I would add nuance is in the composition of value inside software.
AI does not eliminate software. It compresses surface SaaS and expands control layers.
As autonomy increases, execution risk increases. As models move up the stack, governance, permissioning, orchestration, and audit shift from features to infrastructure. The more capable the system, the more valuable the enforcement boundary.
Historically, when capability expanded, constraint layers grew fastest. Payments and fraud grew with e-commerce. Identity and networking grew with distributed systems. CDNs grew with streaming. The enabling constraint layer often becomes the durable economic tier.
The herd conclusion that “software is dead” misses this. We will build more software than ever. The shift is architectural.
This is exactly why I am building ExecLayer, infrastructure that sits between AI intent and real-world execution. If models generate decisions, something must validate authority, permissions, and risk before action is taken. That enforcement layer compounds as capability scales.
The surface will change fast. The control layer will matter more over time.
#ExecLayer✈️ #AgentGovernance
As someone who’s building a GPU-powered Terminal App, the return to a CLI-first experience makes me giggle so hard. The times have changed but only in our memory perhaps! What is old is now new, etc. etc. etc.
I liked this quote but wanted to adjust it a bit:
> “New tools will be created with AI that do new things.”
to:
> “New tools will be created with AI that do old things in new ways.
The gutenberg press enabled writing and publishing (and copy-pasta-ing) at a greater scale. The typewriter did the same and then computer and then the internet and now AI.
Every major leap seems to follow the same pattern:
1. Reduce the cost of reproduction
2. Expand who can produce content
3. Trigger a crisis among existing gatekeepers (church, state, publishers, media companies)
4. Restructure power — those who control information lose monopoly
These are amazing times but our fundamental needs and even workflows really haven’t changed. We are still human after all these years.
Right, so that's why it's time build 🔥
Who is going to do to software what Substack did to media publishing
Make it super easy for me to own and operate a SaaS please
Finally an articulate optimistic perspective with historical evidence. No one knows what’s going to happen, but the idea that jobs are just going to disappear and none will be created is nonsense. At worst, it’ll be a reversion to pre-industrial revolution where most everyone is technically an entrepreneur and has their own company/service. And that’s still pretty awesome!
This tracks with what I'm seeing in the market data. The 'death of software' narrative was always overblown - what's actually happening is consolidation around agent-ready platforms.
The SaaS companies that will survive aren't the ones with the most features. They're the ones that can integrate into agent workflows seamlessly. I looked at 20+ enterprise deployments last week and the pattern is clear: point solutions die, orchestration layers win.
My analysis of which software companies are actually at risk vs. which are positioning to win: https://thoughts.jock.pl/p/ai-agent-landscape-feb-2026-data
The infrastructure spend numbers tell the real story - it's not about killing software, it's about capturing the coordination layer.
Each platform has an optimal AI tool for its formats.
While AI produces output from human input human oversight, imagination, dexterity, creativity, and niche knowledge guarantees the following:
1. Brand voice
2. Accuracy
3. Strategic refinements
Pick the best embedded tool.
I agree — we need more software than ever. But that's not the question. The question is whether the human labor model that built software for fifty years survives when AI can do the implementation. More software, fewer programmers, no sprints. You are defending the industry. Here (it is a series of 6 articles: https://open.substack.com/pub/edgonen/p/directing-the-machine-that-replaced?utm_campaign=post-expanded-share&utm_medium=web) I'm describing what's happening inside it.
AI lowers the cost of coordination, not the demand for outcomes—so SMBs don’t need less software, they need more software that fits their exact workflows.
Under Coasean logic, AI-native technology services exist to redesign firm boundaries: internalizing what’s now cheap (custom logic, coordination) and externalizing only what truly shouldn’t be built in-house.
Great read! AI is a complement to software not a substitute.