Some food for thought as we collectively figure out who (if anyone) is going to have their worst fear realized and be fully replaced by agents.
Those gains haven’t shown up as time saved, though. Time spent on existing tasks in Linear held while AI usage appeared as a new layer of work, meaning the overall time spent on product development is going up rather than down. As far as we can observe, teams are working more, not less, suggesting AI has a Jevons paradox quality beyond token consumption.
Some food for thought as we collectively figure out who (if anyone) is going to have their worst fear realized and be fully replaced by agents.
Executives are personally active on AI at rates that match or beat their teams. CEOs at companies of 201 or more people went from 9% to 36% in six months, the largest jump of any cut in this report, suggesting the most senior leaders are learning the technology by using it rather than reading about it. Company size comes from third-party enrichment, so this cut covers fewer workspaces than the rest of the report.
[...]
Two years ago, fewer than one issue in a thousand was created by AI. Teams now use AI to write just under half of everything created in Linear, and at the current pace it will soon author more than people and integrations combined.
[...]
Teams that connected a coding agent roughly tripled their weekly pull requests over two years, from 21 to 65, while teams without one went from 8 to 10. These teams were already higher-output before coding agents existed, so the levels aren’t directly comparable, but each cohort against its own baseline tells a clean story, and nearly all the growth sits on the agent side.
[...]
The clearest indication of AI’s influence on product development is the dramatic output gains experienced by teams using coding agents over the last two years. We have no way of knowing whether this increased output led to positive business outcomes, but it shows a very clear correlation between AI adoption and acceleration.
Perhaps more intriguing is the makeup of that adoption, and how it appears to be blurring roles. Senior leaders are doing more of the hands-on IC work, adopting AI aggressively to help them do it, and non-engineers are committing code. The suggestion that everyone in an organization is becoming a “builder” seems to be directionally true.
Those gains haven’t shown up as time saved, though. Time spent on existing tasks in Linear held while AI usage appeared as a new layer of work, meaning the overall time spent on product development is going up rather than down. As far as we can observe, teams are working more, not less, suggesting AI has a Jevons paradox quality beyond token consumption.
Many will rightfully argue that looking at pull requests indicates motion rather than value, which is certainly true, but it’s still a step forward from measuring tokens. A mechanical refactor might burn lots of tokens while a meaningful bug fix or code review doesn’t, so token spend and value don’t line up at all, and using one as a proxy for the other will be remembered as a relic of AI’s early days.
We've been doing this for a while at my current job. Importantly, we are reading and editing our cards before marking them as "Ready" though I see the time spent on this to be decreasing. I think...
Two years ago, fewer than one issue in a thousand was created by AI. Teams now use AI to write just under half of everything created in Linear, and at the current pace it will soon author more than people and integrations combined.
We've been doing this for a while at my current job. Importantly, we are reading and editing our cards before marking them as "Ready" though I see the time spent on this to be decreasing. I think something Sean Goedecke mentioned recently is playing a part here.
It does occasionally worry me that more and more of the "card lifetime" is being handled by LLMs. For us, it seems like there's a great deal of care that goes into triaging an issue or a feature (still LLM-assisted). I mentioned this a few months ago in another thread. That still holds true. Since then, the bias towards the LLM almost owning the process of development and review (and these days, significantly assisting me in my testing, usually through harnesses and I/O validation) has continued to dominate. There's not a single part of my workflow where I'm not using an LLM to do something. I think I still have mixed feelings about that.
On one hand, it's nice that LLM models are getting to the point where a trivial, bit-sized piece of work is almost entirely automated where my role is essentially guiding it to completion. I have personally found it to be an excellent and efficient tool for performing a guided task. On the other hand, I used to spend a lot more time actually writing code. Sometimes I would even enjoy the process. Other times, if I didn't, I would at least feel like I had ownership of it. That is less so the case these days.
For my personal projects, I get to enjoy the best of both worlds. At work, the cost-benefit of using LLM models is that we're able to move quickly. I can slow down for my personal projects and actually write the code that is enjoyable to write. I can let an LLM do the parts I would usually anguish over with very little gained at the end. I have much more fun coding on personal projects these days than I did pre-LLM. I can basically set my own difficulty level now.
Just got to the end of writing this and realized it was more than I expected to write, so if you made it this far, thank you for participating in my train of thought. Every so often I like to hear how others in the industry feel about the current state of their jobs with respect to LLMs. My feeling as of writing this is... bittersweet. Usually I'm a little more excited about it LOL
I think you managed to encapsulate exactly how I feel about AI-assisted coding. So I'll just add some meta points The current trend give you a ton of job insecurity. It is my belief that both labs...
I think you managed to encapsulate exactly how I feel about AI-assisted coding. So I'll just add some meta points
The current trend give you a ton of job insecurity. It is my belief that both labs and executives are aligned to remove or rather eliminate at least 50% of engineering roles (let's just say medium skill and less). I worry a lot more about the future of the field. Before I would encourage and sometimes even strongly nudge a friend to learn coding. Now I'm not so sure.
Creativity, craft, and quality software will continue to degrade than the current average. Enshittification is already a thing and AI wil only make it worse. This is because most companies and their execs don't care for those. So software will be functional for sure but they won't feel good to use. Heck you'll probably have so many features that you're not even sure where to click. Everything will feel like Zoom. Ugh.
In addition to what you enjoy on your personal projects, I really enjoy crafting the design and UX and then perfecting it to pixel perfect. As you said there is no deadline on personal projects so taking a week to make the app more performant or to experiment with UX feels like bliss meditation.
Yea I have a lot of mixed feelings as well. Fortunately I work on an interesting product/project, so I still get enjoyment from making it better. But I can imagine if you were working on e.g a...
Yea I have a lot of mixed feelings as well. Fortunately I work on an interesting product/project, so I still get enjoyment from making it better. But I can imagine if you were working on e.g a backend system for a generic crud app your life just becomes a wall of LLM assissted ticket processing and there’s no longer any “fun coding” parts to take
solace in.
Indeed. I have a good amount of friends in startups and its just 100% vibecoding without much restraint. All kind of dumb stuff are built, very little thought into architecture. UX or any redesign...
Indeed. I have a good amount of friends in startups and its just 100% vibecoding without much restraint. All kind of dumb stuff are built, very little thought into architecture. UX or any redesign is infinitely delayed until "next quarter". It's really a shit show. I don't think its just startups, I think its happening in most companies. It's just that understandably pretty much nobody dares to talk about it without backlash.
It's another reason why I always suggest to friends to do a very long and complex side project if time/life permits.
Some food for thought as we collectively figure out who (if anyone) is going to have their worst fear realized and be fully replaced by agents.
From the article:
[...]
[...]
[...]
We've been doing this for a while at my current job. Importantly, we are reading and editing our cards before marking them as "Ready" though I see the time spent on this to be decreasing. I think something Sean Goedecke mentioned recently is playing a part here.
It does occasionally worry me that more and more of the "card lifetime" is being handled by LLMs. For us, it seems like there's a great deal of care that goes into triaging an issue or a feature (still LLM-assisted). I mentioned this a few months ago in another thread. That still holds true. Since then, the bias towards the LLM almost owning the process of development and review (and these days, significantly assisting me in my testing, usually through harnesses and I/O validation) has continued to dominate. There's not a single part of my workflow where I'm not using an LLM to do something. I think I still have mixed feelings about that.
On one hand, it's nice that LLM models are getting to the point where a trivial, bit-sized piece of work is almost entirely automated where my role is essentially guiding it to completion. I have personally found it to be an excellent and efficient tool for performing a guided task. On the other hand, I used to spend a lot more time actually writing code. Sometimes I would even enjoy the process. Other times, if I didn't, I would at least feel like I had ownership of it. That is less so the case these days.
For my personal projects, I get to enjoy the best of both worlds. At work, the cost-benefit of using LLM models is that we're able to move quickly. I can slow down for my personal projects and actually write the code that is enjoyable to write. I can let an LLM do the parts I would usually anguish over with very little gained at the end. I have much more fun coding on personal projects these days than I did pre-LLM. I can basically set my own difficulty level now.
Just got to the end of writing this and realized it was more than I expected to write, so if you made it this far, thank you for participating in my train of thought. Every so often I like to hear how others in the industry feel about the current state of their jobs with respect to LLMs. My feeling as of writing this is... bittersweet. Usually I'm a little more excited about it LOL
I think you managed to encapsulate exactly how I feel about AI-assisted coding. So I'll just add some meta points
The current trend give you a ton of job insecurity. It is my belief that both labs and executives are aligned to remove or rather eliminate at least 50% of engineering roles (let's just say medium skill and less). I worry a lot more about the future of the field. Before I would encourage and sometimes even strongly nudge a friend to learn coding. Now I'm not so sure.
Creativity, craft, and quality software will continue to degrade than the current average. Enshittification is already a thing and AI wil only make it worse. This is because most companies and their execs don't care for those. So software will be functional for sure but they won't feel good to use. Heck you'll probably have so many features that you're not even sure where to click. Everything will feel like Zoom. Ugh.
In addition to what you enjoy on your personal projects, I really enjoy crafting the design and UX and then perfecting it to pixel perfect. As you said there is no deadline on personal projects so taking a week to make the app more performant or to experiment with UX feels like bliss meditation.
Yea I have a lot of mixed feelings as well. Fortunately I work on an interesting product/project, so I still get enjoyment from making it better. But I can imagine if you were working on e.g a backend system for a generic crud app your life just becomes a wall of LLM assissted ticket processing and there’s no longer any “fun coding” parts to take
solace in.
Indeed. I have a good amount of friends in startups and its just 100% vibecoding without much restraint. All kind of dumb stuff are built, very little thought into architecture. UX or any redesign is infinitely delayed until "next quarter". It's really a shit show. I don't think its just startups, I think its happening in most companies. It's just that understandably pretty much nobody dares to talk about it without backlash.
It's another reason why I always suggest to friends to do a very long and complex side project if time/life permits.