I recently read a tweet from the famed indie dev levelsio about how indie dev revenue is declining. As an indie musician and an entrepreneur, I am sympathetic.
Remember Napster?
I was 18 and actively playing in my metal band when Metallica sued Napster for leaking ‘I Disappear’ (part of the Mission: Impossible II soundtrack) on their P2P sharing platform. Metallica took a lot of flak for punishing young fans who were ‘sharing their music’.
No matter what side of the argument you found yourself on, it was Napster that led to torrents, and eventually to Spotify.
Almost 25 years later, it is argued that Spotify and later other Digital Streaming Platforms (DSPs) led to the demise of independent music as a viable profession, and made it almost impossible to make a living off of music. We had iTunes and Zune before Spotify, but digital downloads and streaming are entirely different business models. But what they share in common is that the supply of music now outweighs its demand. If not as an absolute, then at least at the moment of discovery.
“In an information-rich world, the wealth of information means a dearth of something else: a scarcity of whatever it is that information consumes. What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention.”
— Herbert Simon, a Nobel Prize–winning social scientist
The economics flipped from scarcity of the good to scarcity of attention. A listener’s attention is finite: supply can grow without limit, but listening hours cannot. Fixed demand met exploding supply. A horrible arithmetic for musicians.
Despite this, streaming didn’t kill the industry (total industry revenue is back to 90s levels); it just had your local artist competing with their idols for attention, mediated by a platform’s playlist algorithm. The music survived. The musicians’ economics did not.
Software has a platform problem.
With AI, the cost of producing and distributing the good (software) is collapsing to zero, the same way streaming collapsed the cost of distributing your music. And here’s my claim: software is entering its Spotify era. However, since software and music don’t share the same demand curve, the fight will not be over attention but over revenue capture.
While everyone’s focused on AI as a production mechanism, I am interested in its distribution mechanism. There are two facets:
AI chatbots as discovery mechanism: For years, you found software on an app store or searched Google for it. However, search traffic is declining, and chatbots are becoming the primary surface where users research their options. Web traffic is declining with it. Another indie dev, Jon Yongfook, complained that it is all downstream from website traffic and that’s what is drying up.
Coding agents as production + discovery mechanism: The more interesting surface is the coding agent. This is a step change in how users build and discover solutions to their problems. Users don’t go about thinking of ways to buy software; they interact with an agent to solve a problem. And these agents solve problems before software, as a product to be discovered and bought, ever needs to happen.
So, is there a Silver Lining?
Software and music don’t share the same demand curve. Software doesn’t consume attention; it consumes jobs-to-be-done. And new software keeps creating new jobs-to-be-done. Verifying AI’s output before it is integrated is a job that didn’t exist until recently.
This also means that markets that couldn’t be served with software before are serviceable now. Small businesses that needed one specific piece of their operation automated can finally get it.
And, in software, ‘good enough’ is often good enough. The distaste we feel about AI slop in music or arts doesn’t apply to jobs that were never interesting enough for professional engineers to solve. That means more demand, not less.
What’s the catch?
Elastic demand does not mean elastic revenue capture. The total addressable market for “custom software solving small problems” may be bigger than ever before, but who captures it is a different question. Right now, the AI tooling providers are positioned to capture it, much like DSPs in music. They own the surface where discovery happens — both problem and solution in most cases. The playlist algorithm has a successor: whatever repo or code snippet the agent decides to reach for.
In Conclusion
The economics of indie software are being rewritten in real time. In less than three months, we went from a burst of solo founder companies run by a thousand agents to the return of teams — the rise of multiplayer AI.
I am not a betting man. But if I were, this time I’d place one: the demand survives, the makers multiply, and the platforms take the margin.
In music, we learned that too late.
title: 'The Spotify Era of Software' publishedAt: '2026-07-29' summary: 'What happens when everyone can make software' tags: ['ai', 'software', 'indie']
I recently read a tweet from the famed indie dev levelsio about how indie dev revenue is declining. As an indie musician and an entrepreneur, I am sympathetic.
Remember Napster?
I was 18 and actively playing in my metal band when Metallica sued Napster for leaking 'I Disappear' (part of the Mission: Impossible II soundtrack) on their P2P sharing platform. Metallica took a lot of flak for punishing young fans who were 'sharing their music'.
No matter what side of the argument you found yourself on, it was Napster that led to torrents, and eventually to Spotify.
Almost 25 years later, it is argued that Spotify and later other Digital Streaming Platforms (DSPs) led to the demise of independent music as a viable profession, and made it almost impossible to make a living off of music.
We had iTunes and Zune before Spotify, but digital downloads and streaming are entirely different business models. But what they share in common is that the supply of music now outweighs its demand. If not as an absolute, then at least at the moment of discovery.
"In an information-rich world, the wealth of information means a dearth of something else: a scarcity of whatever it is that information consumes. What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention." — Herbert Simon, a Nobel Prize–winning social scientist
The economics flipped from scarcity of the good to scarcity of attention. A listener's attention is finite: supply can grow without limit, but listening hours cannot. Fixed demand met exploding supply. A horrible arithmetic for musicians.
Despite this, streaming didn't kill the industry (total industry revenue is back to 90s levels); it just had your local artist competing with their idols for attention, mediated by a platform's playlist algorithm. The music survived. The musicians' economics did not.
Software has a platform problem.
With AI, the cost of producing and distributing the good (software) is collapsing to zero, the same way streaming collapsed the cost of distributing your music. And here's my claim: software is entering its Spotify era. However, since software and music don't share the same demand curve, the fight will not be over attention but over revenue capture.
While everyone's focused on AI as a production mechanism, I am interested in its distribution mechanism. There are two facets:
AI chatbots as discovery mechanism: For years, you found software on an app store or searched Google for it. However, search traffic is declining, and chatbots are becoming the primary surface where users research their options. Web traffic is declining with it. Another indie dev, Jon Yongfook, complained that it is all downstream from website traffic and that's what is drying up.
Coding agents as production + discovery mechanism: The more interesting surface is the coding agent. This is a step change in how users build and discover solutions to their problems. Users don't go about thinking of ways to buy software; they interact with an agent to solve a problem. And these agents solve problems before software, as a product to be discovered and bought, ever needs to happen.
So, is there a Silver Lining?
Software and music don't share the same demand curve. Software doesn't consume attention; it consumes jobs-to-be-done. And new software keeps creating new jobs-to-be-done. Verifying AI's output before it is integrated is a job that didn't exist until recently.
This also means that markets that couldn't be served with software before are serviceable now. Small businesses that needed one specific piece of their operation automated can finally get it.
And, in software, 'good enough' is often good enough. The distaste we feel about AI slop in music or arts doesn't apply to jobs that were never interesting enough for professional engineers to solve. That means more demand, not less.
What's the catch?
Elastic demand does not mean elastic revenue capture. The total addressable market for "custom software solving small problems" may be bigger than ever before, but who captures it is a different question. Right now, the AI tooling providers are positioned to capture it, much like DSPs in music. They own the surface where discovery happens — both problem and solution in most cases. The playlist algorithm has a successor: whatever repo or code snippet the agent decides to reach for.
In Conclusion
The economics of indie software are being rewritten in real time. In less than three months, we went from a burst of solo founder companies run by a thousand agents to the return of teams — the rise of multiplayer AI.
I am not a betting man. But if I were, this time I'd place one: the demand survives, the makers multiply, and the platforms take the margin.
In music, we learned that too late.