N
Nica Furs
Guest
For most of the history of software, the developer was the bottleneck. Code moved as fast as someone could read, reason, and type. Agentic coding tools changed that arrangement. In Claude Code, Cursor, and Codex, a developer now describes a task, the agent goes off to read the repository, run commands, and propose changes, and the developer waits for the result.
The individual waits are not long. A focused refactor might take under a minute, and a change that touches several files might take a few. But they happen dozens of times a day, and together they have created a new pattern in how developers spend their working hours: short, repeated stretches where their eyes are on the screen and their hands are off the keyboard.
That pattern is worth looking at closely, because almost every major attention market in the last century started with the same thing: a new behavior that produced idle, predictable moments.
Commercial radio sold the commute. Television sold the evening. Search engines sold the moment of intent between a question and an answer. Social feeds sold the idle scroll. In each case, the inventory was valuable for three reasons. The moment was predictable, the audience was identifiable, and the attention could be measured well enough for an advertiser to pay for it.
AI coding wait time has all three properties.
It is predictable because it is triggered by a discrete event. The coding tool knows exactly when an agent starts working and when it stops, so the window can be detected without guesswork.
The audience is identifiable because the people using frontier coding agents are, almost by definition, technical decision-makers. They choose cloud providers, APIs, databases, developer tools, and increasingly the financial infrastructure their products run on. That is one of the most valuable audiences in technology, and historically one of the hardest to reach.
The attention is measurable because the message appears in the same terminal or editor the developer is already watching while waiting for the agent to finish.
There is also a property that most ad inventory lacks. A pre-roll video stands between a viewer and what they came for. A message shown while an agent is working does not delay anything, because the agent is running regardless. The developer is not interrupted; they are simply idle.
Developer marketing has long relied on indirect channels: documentation, conference sponsorships, newsletter placements, open-source sponsorships, and free credits. Display advertising performs poorly with this audience, partly because ad-blocker use among developers is high and partly because developers are trained to ignore anything that looks like a banner.
The places developers spend most of their time, the editor and the terminal, have been off-limits to advertising, and for good reason. They are workspaces, and anything that competes for attention inside a workspace has a cost in focus.
Agents changed the shape of those workspaces. They inserted idle intervals directly inside the environment where developers work. The question is no longer whether there is attention available inside the editor. It is whether that attention can be used without degrading the work around it, and who should be paid for it.
In conventional advertising, the value of attention flows to the platform and the publisher. The person whose attention is being sold receives a free service at best.
Wait-time attention inside a developer's own machine breaks that model, because there is no third-party publisher. The developer owns the screen, the session, and the idle time. If that attention is sold, the logical recipient of the payment is the developer.
The idea of paying users directly for ad attention is not new. Brave has offered users rewards in its own token for opting into ads since 2019. What is new is the surface. Coding agents create a high-value, well-defined window that did not exist two years ago.
One of the first products built around this window is Relo, an ad network that shows a single sponsored line while Claude Code, Cursor, or Codex is working and pays the developer for viewing it. Its design illustrates the decisions any product in this category has to make.
Installation is a single command, npx relo-money init, which configures the coding tools it finds on the machine. In Claude Code and Codex, the sponsored line appears in the terminal while the agent runs. In Cursor, a lightweight extension shows it in the status bar. When the agent starts working, one sponsored message appears. If it stays visible for 10 continuous seconds, it counts as a verified view.
Advertisers pay only for verified views, and the advertiser-funded reward is added to the developer's balance. Campaigns can pay out in project tokens, stablecoins, or tokenized AI stocks, and developers claim rewards directly to their own wallet. Every ad is reviewed before it runs, and advertisers confirm the campaign budget and reward asset in advance.
The choice of reward assets reflects the audience. A developer who spends the day watching AI agents work is likely to find equity exposure to the companies building those agents more interesting than ad-network points.
Any software that runs inside a coding environment raises an obvious concern. The editor and the terminal contain proprietary code, credentials, and conversations with AI tools about unreleased work. An ad network that touched any of that would be unacceptable to most developers and most employers.
This is where the category will be won or lost, and it is the area where trust has to be earned through transparency rather than marketing claims. Relo's approach is to publish its client as a source-available mirror on GitHub so developers can read exactly what runs on their machine. According to that repository, the client communicates only with Relo's own server and sends four things: a device ID created at setup, the payout wallet address the developer enters, which tool started or stopped, and the ID of the sponsored line once it has stayed up long enough to count.
The repository also addresses a subtler issue. Coding tool hooks can pass the user's prompt to the client through standard input. Relo's client reads that input only so the process does not stall, then discards it without storing or sending it. The client does not read code, files, or AI conversations, and the Cursor extension reads only Relo's own local state file.
The code is source-available rather than open source, which means developers can audit it but cannot fork or redistribute it. For a product whose value depends on trust, publishing the client is a reasonable baseline, and it is likely the minimum standard that developers will expect from anyone building in this space.
The category is new, and several questions will determine whether it becomes a durable market or a curiosity.
The first is acceptance. Developers have a strong and justified aversion to ads in their tools. Making the experience opt-in, limiting it to one line during idle time, and paying the developer directly changes the calculation, but whether it changes it enough at scale remains to be seen.
The second is whether the wait will last. Models are getting faster, which could shrink individual waits. At the same time, agents are being given larger tasks and longer autonomous runs, so the total time developers spend watching agents work may shift rather than disappear.
The third is verification. Any network that pays users for attention attracts people who try to fake that attention. A 10-second visibility rule is a sensible starting point, but resistance to automated and duplicate accounts will ultimately decide whether advertisers trust the inventory.
The fourth is the position of the tool makers themselves. The companies behind coding agents control the surfaces where this attention lives, and they may choose to restrict third-party integrations or build monetization of their own.
Attention markets are rarely invented. They appear when technology changes how people spend their time and someone notices that a new, measurable moment has emerged. AI coding agents have created one of those moments inside the most guarded workspace in technology.
Whether it becomes a meaningful market depends less on the technology than on the terms. Developers will accept a sponsored line in their terminal only if it stays out of their work, respects their code, and pays them fairly for their time. The products that meet those conditions will define how this new market operates.
This article was published under HackerNoon's Business Blogging program.
The individual waits are not long. A focused refactor might take under a minute, and a change that touches several files might take a few. But they happen dozens of times a day, and together they have created a new pattern in how developers spend their working hours: short, repeated stretches where their eyes are on the screen and their hands are off the keyboard.
That pattern is worth looking at closely, because almost every major attention market in the last century started with the same thing: a new behavior that produced idle, predictable moments.
Every Attention Market Starts With a New Kind of Waiting
Commercial radio sold the commute. Television sold the evening. Search engines sold the moment of intent between a question and an answer. Social feeds sold the idle scroll. In each case, the inventory was valuable for three reasons. The moment was predictable, the audience was identifiable, and the attention could be measured well enough for an advertiser to pay for it.
AI coding wait time has all three properties.
It is predictable because it is triggered by a discrete event. The coding tool knows exactly when an agent starts working and when it stops, so the window can be detected without guesswork.
The audience is identifiable because the people using frontier coding agents are, almost by definition, technical decision-makers. They choose cloud providers, APIs, databases, developer tools, and increasingly the financial infrastructure their products run on. That is one of the most valuable audiences in technology, and historically one of the hardest to reach.
The attention is measurable because the message appears in the same terminal or editor the developer is already watching while waiting for the agent to finish.
There is also a property that most ad inventory lacks. A pre-roll video stands between a viewer and what they came for. A message shown while an agent is working does not delay anything, because the agent is running regardless. The developer is not interrupted; they are simply idle.
Why Developers Have Been So Hard to Reach
Developer marketing has long relied on indirect channels: documentation, conference sponsorships, newsletter placements, open-source sponsorships, and free credits. Display advertising performs poorly with this audience, partly because ad-blocker use among developers is high and partly because developers are trained to ignore anything that looks like a banner.
The places developers spend most of their time, the editor and the terminal, have been off-limits to advertising, and for good reason. They are workspaces, and anything that competes for attention inside a workspace has a cost in focus.
Agents changed the shape of those workspaces. They inserted idle intervals directly inside the environment where developers work. The question is no longer whether there is attention available inside the editor. It is whether that attention can be used without degrading the work around it, and who should be paid for it.
Paying the Viewer, Not Just the Platform
In conventional advertising, the value of attention flows to the platform and the publisher. The person whose attention is being sold receives a free service at best.
Wait-time attention inside a developer's own machine breaks that model, because there is no third-party publisher. The developer owns the screen, the session, and the idle time. If that attention is sold, the logical recipient of the payment is the developer.
The idea of paying users directly for ad attention is not new. Brave has offered users rewards in its own token for opting into ads since 2019. What is new is the surface. Coding agents create a high-value, well-defined window that did not exist two years ago.
One of the first products built around this window is Relo, an ad network that shows a single sponsored line while Claude Code, Cursor, or Codex is working and pays the developer for viewing it. Its design illustrates the decisions any product in this category has to make.
Installation is a single command, npx relo-money init, which configures the coding tools it finds on the machine. In Claude Code and Codex, the sponsored line appears in the terminal while the agent runs. In Cursor, a lightweight extension shows it in the status bar. When the agent starts working, one sponsored message appears. If it stays visible for 10 continuous seconds, it counts as a verified view.
Advertisers pay only for verified views, and the advertiser-funded reward is added to the developer's balance. Campaigns can pay out in project tokens, stablecoins, or tokenized AI stocks, and developers claim rewards directly to their own wallet. Every ad is reviewed before it runs, and advertisers confirm the campaign budget and reward asset in advance.
The choice of reward assets reflects the audience. A developer who spends the day watching AI agents work is likely to find equity exposure to the companies building those agents more interesting than ad-network points.
Privacy Is the Product
Any software that runs inside a coding environment raises an obvious concern. The editor and the terminal contain proprietary code, credentials, and conversations with AI tools about unreleased work. An ad network that touched any of that would be unacceptable to most developers and most employers.
This is where the category will be won or lost, and it is the area where trust has to be earned through transparency rather than marketing claims. Relo's approach is to publish its client as a source-available mirror on GitHub so developers can read exactly what runs on their machine. According to that repository, the client communicates only with Relo's own server and sends four things: a device ID created at setup, the payout wallet address the developer enters, which tool started or stopped, and the ID of the sponsored line once it has stayed up long enough to count.
The repository also addresses a subtler issue. Coding tool hooks can pass the user's prompt to the client through standard input. Relo's client reads that input only so the process does not stall, then discards it without storing or sending it. The client does not read code, files, or AI conversations, and the Cursor extension reads only Relo's own local state file.
The code is source-available rather than open source, which means developers can audit it but cannot fork or redistribute it. For a product whose value depends on trust, publishing the client is a reasonable baseline, and it is likely the minimum standard that developers will expect from anyone building in this space.
The Open Questions
The category is new, and several questions will determine whether it becomes a durable market or a curiosity.
The first is acceptance. Developers have a strong and justified aversion to ads in their tools. Making the experience opt-in, limiting it to one line during idle time, and paying the developer directly changes the calculation, but whether it changes it enough at scale remains to be seen.
The second is whether the wait will last. Models are getting faster, which could shrink individual waits. At the same time, agents are being given larger tasks and longer autonomous runs, so the total time developers spend watching agents work may shift rather than disappear.
The third is verification. Any network that pays users for attention attracts people who try to fake that attention. A 10-second visibility rule is a sensible starting point, but resistance to automated and duplicate accounts will ultimately decide whether advertisers trust the inventory.
The fourth is the position of the tool makers themselves. The companies behind coding agents control the surfaces where this attention lives, and they may choose to restrict third-party integrations or build monetization of their own.
A Market Created by a Change in Behavior
Attention markets are rarely invented. They appear when technology changes how people spend their time and someone notices that a new, measurable moment has emerged. AI coding agents have created one of those moments inside the most guarded workspace in technology.
Whether it becomes a meaningful market depends less on the technology than on the terms. Developers will accept a sponsored line in their terminal only if it stays out of their work, respects their code, and pays them fairly for their time. The products that meet those conditions will define how this new market operates.
This article was published under HackerNoon's Business Blogging program.