Well as a person who is working as a software developer I wouldn’t be so hasty.
You can write more code, but that has never been a real bottleneck. Understanding and maintenance of this code is another matter altogether.
Add to that the price of AI subscriptions are currently heavily subsidized by venture capital and even with the subsidies tokens turn out to be more expensive than people.
Where are AI subscriptions subsidised for enterprise use? Github copilot was the last to drop the subsidised model for big business at the start of the month as far as I can tell. Only individuals and very small businesses are getting subsidised subscriptions now, and it’s still super economical and cost efficient to use even frontier models at API billing rates compared to humans. A human can work all day on debugging a software defect, or Opus can find the root cause in ten minutes for $20. Sure that still needs reviewing but that’s insane productivity AND cost improvement
A human can work all day on debugging a software defect, or Opus can find the root cause in ten minutes for $20
Yeah or it can delete your prod database without asking you.
Additionally the heavy use of AI can lead to comprehension debt meaning no one can understand it. AI is good if it has the data but usually the data is not only code it’s Kafka and infrastructure and other ongoing outages that may be related and logs.
This is more of a parallel of the video game crash that happened before. When the video game consoles created a bubble in the US every body suddenly started creating video games, to the point many were so bad they were literally unplayable. When the market got flooded with bad games, people stopped buying games (since no one trusted the quality anymore), leading to a crash in the industry: https://en.wikipedia.org/wiki/Video_game_crash_of_1983
For my projects, finding the cause of a bug is rarely a problem once it’s reported. It’s fixing it in a way that doesn’t negatively impact things upstream or downstream that’s a pain.
How’s AI supposed to help when we’ve got to negotiate with several other stakeholders on what changes we’re going to make?
Well as a person who is working as a software developer I wouldn’t be so hasty.
You can write more code, but that has never been a real bottleneck. Understanding and maintenance of this code is another matter altogether.
Add to that the price of AI subscriptions are currently heavily subsidized by venture capital and even with the subsidies tokens turn out to be more expensive than people.
Also no one is calling it SaaS apocalypse.
https://www.economist.com/business/2026/06/10/fear-of-the-saaspocalypse-is-tormenting-techland
https://www.forbes.com/sites/donmuir/2026/02/04/300-billion-evaporated-the-saaspocalypse-has-begun/
https://techcrunch.com/2026/03/01/saas-in-saas-out-heres-whats-driving-the-saaspocalypse/
https://www.risingtrends.co/blog/saas-apocalypse-trend
It is very much being called the SaaS apocalypse…
Where are AI subscriptions subsidised for enterprise use? Github copilot was the last to drop the subsidised model for big business at the start of the month as far as I can tell. Only individuals and very small businesses are getting subsidised subscriptions now, and it’s still super economical and cost efficient to use even frontier models at API billing rates compared to humans. A human can work all day on debugging a software defect, or Opus can find the root cause in ten minutes for $20. Sure that still needs reviewing but that’s insane productivity AND cost improvement
How is not a single AI company profitable then?
No
Yeah or it can delete your prod database without asking you. Additionally the heavy use of AI can lead to comprehension debt meaning no one can understand it. AI is good if it has the data but usually the data is not only code it’s Kafka and infrastructure and other ongoing outages that may be related and logs.
This is more of a parallel of the video game crash that happened before. When the video game consoles created a bubble in the US every body suddenly started creating video games, to the point many were so bad they were literally unplayable. When the market got flooded with bad games, people stopped buying games (since no one trusted the quality anymore), leading to a crash in the industry: https://en.wikipedia.org/wiki/Video_game_crash_of_1983
For my projects, finding the cause of a bug is rarely a problem once it’s reported. It’s fixing it in a way that doesn’t negatively impact things upstream or downstream that’s a pain.
How’s AI supposed to help when we’ve got to negotiate with several other stakeholders on what changes we’re going to make?