Scott Carleton

Lover of iteration and continuous learning. Formerly Asana, Andela, TpT, Artsicle

[Dan Shipper 📧 @danshipper]

when you experience a language model doing a task that used to require you at every step and now doesn't, it's a gigantic agency rupture.

we tend to equate ourselves with our tasks, our outputs, and a rupture of this kind is akin to the loss of your identity and sense of self.

you are no longer required, and you had no say in the matter. it's a kind of death.

however, agency ruptures with AI seem to follow a predictable pattern:

- Initial agency rupture. Here you see the AI doing what you used to do, and you see only the AI. In your perception of reality the AI is highlighted, and the human work around getting the AI to do the thing, is invisible. you can see this in the way we use language: "AI solved XYZ Erdos Problem" is another way to say we're in the initial rupture phase.

- Seeing human scaffolding. After a little bit of experience with the model and its new capibility—programming, writing, math, etc—we start to see the edges of what's possible with the model. Here, instead of just seeing the AI doing the task, we start to see the scaffolding required by you and other humans on either end to make sure the model does its work. At this stage, the scaffolding can feel like a secondary part of the work, but it is work nonetheless. We begin to say things like, "The AI is prompting me", or "My job is just to babysit Claude."

- Agency reconstruction. After a long enough time with the model at its current level of capability, we start to reconstruct a new sense of agency that centers our own role in the work being done, and makes the model's contribution mostly invisible. It is now a tool, it fast-forwards you through the boring or repetitive parts of the work, and what used to seem like scaffolding now seems like the actual work itself—the interesting part. You've had enough experience with the scaffolding to see its nuances, and how hard it is to get the model to reliably output high quality work. (As part of this, your conception of what quality work is changes—the floor is raised, and so is the ceiling.)

You can tell you're at this point because you stop saying "The AI did this" and just say, "I did this" and the fact that you used AI is implied. For example, it would be weird for me to talk to someone @every and ask them if the AI built the most recent feature of a product. Of course it did! But we're so used to the capability that it becomes invisible, and we re-center ourselves and our own conception of agency.

My theory is that the ability to metabolize agency ruptures and turn them into playfulness, and curiosity is a good leading indicator for whether you're a person who likes and excels in the new AI economy or not.

Each time capabilities jump, you get new ruptures for existing fields touched by AI. And you get new ruptures for naive populations (like mathemeticians) who until recently hadn't been affected.

if you know what the cycle looks like, it makes it much easier to ride it without freaking out. i think most people @every are naturally good at this

[Scott Carleton @ScotterC]

@aviflombaum Unbelievable. I can’t stop thinking about how great that was

[Dan Shipper 📧 @danshipper]

We’ve automated every single thing we can @every with AI agents.

And yet there’s way more human work to do than ever. We’ve gone from 4 -> 30 human employees since GPT-3.

I wrote a report on the structural reasons: how AI makes expert competence cheap, why that drives up demand for experts, and why the dynamic only intensifies as we approach AGI.

After Automation: https://t.co/Lb7SUCduAg

[Scott Carleton @ScotterC]

@aviflombaum It’s certainly a bit of a shock. I wonder if I’ll code by hand when I’m older as a form of wood working or something.

[Scott Carleton @ScotterC]

@OpenAI seems to be covered by https://t.co/k8kDccO5X4

@openai Measuring fine tune response quality would be worthwhile. This is not the first time we've felt unsupported on core workflows that we depend on OpenAI for

[Scott Carleton @ScotterC]

. @OpenAI our FT models seemed to 404 at an increasing rate all of a sudden
11/17/25 - we had 108k FT requests and 107 errors 11/18/25 - we had 96k FT requests and 32k were errors 11/19/25 - we have 73k FT requests so far and 57k are error

[Scott Carleton @ScotterC]

@_st0012 Would love to use it

[Scott Carleton @ScotterC]

@kasratweets Awesome. I didn’t know anyone who actually read this. I was considering it after a “strange loop” which alone made a great argument of analogies as the basis of thought.

[Scott Carleton @ScotterC]

@nateberkopec Especially with reasoning on tap. Lower cost of reasoning => higher demand on it => more complexity to be managed as we create more.

[Scott Carleton @ScotterC]

@albertwenger I've been thinking along the same lines for a few years. Taken the approach of focusing on 'knowledge extraction' of experts because a business model was required to provide an engine.

[Scott Carleton @ScotterC]

@tobi @factoriogame This is absolute gold. Though I can’t imagine letting my kids start hard drugs so early ;)

[Yishan @yishan]

Moral hazard or not, I now feel the climate situation is bad enough that we should begin scalability work for stratospheric aerosol injection (SAI) immediately.

This is not the same conclusion I would have had even two years ago, but the increase in ocean temperature and extreme climate events indicates a trend that will rapidly get worse unless we are able to take global-scale action within the next 1-5 years, and SAI is the only feasible one.

For those whose initial reaction is opposed, there are a few key things you should be aware of:

- One common fear is that this will be bad for crop yields. I thought this too, but the existing data from volcanic eruptions (which have similar effect) indicated a neutral to positive (!) productivity effect on crops.

- This is not "polluting the air with sulphur." The amount of SO2 needed to significantly induce cooling is on the order of 1% of the SO2 pollution we currently emit, and we would be injecting it into the upper atmosphere. Existing SO2 pollution occurs much lower down, so moving it much higher would likely be better, in terms of health/pollution effects.

- The cessation of sulphur emissions from ships since the 2020 ban on those fuels has given us strong evidence that the prior SO2 emitted by those ships had an (unintended) anti-warming effect on the Atlantic shipping lanes, which is now warming rapidly. While it was also unhealthy pollution, it gives us strong real-world data that this would work at large scale, and we can do it without the harmful pollution side effects by injecting it in the higher atmosphere.

At this point I believe the facts now this conclusion should be relatively uncontroversial if one is practical about looking for solutions.

I am the "tree guy" and in 2020 I would not have supported this, as I felt the world could move quickly to a large-scale reforestation and land restoration effort to make significant progress by 2030. But pandemic, wars, and recession have prevented this (along with good ol' inertia), and warming has accelerated.

Would successful implementation of SAI reduce incentive to move away from fossil fuels? It is a very real risk, yes. In fact, I personally think it is likely.

But the hard brutal reality is that the heating trends right now are very dire, and immediate action to reduce the heating are necessary.

We must begin scaling SAI immediately precisely so that things like reforestation and other carbon capture solutions have time for implementation, which in turn buys time for decarbonization of our economies.

[Scott Carleton @ScotterC]

@kasratweets Discipline ends up being a path to mental freedom. Feeling anxiety about your health? Create a disciplined habit to move more and eat less and you become free of that fear.

[Scott Carleton @ScotterC]

@daniellellecco Oh. I said fun. Which it is. But it’s also dark. Working at the edge of knowledge is apparently very lonely and crazy inducing

[Scott Carleton @ScotterC]

@daniellellecco it's a fun read. the line between fiction and non-fiction gets blurred in a fun way

[Scott Carleton @ScotterC]

@nateberkopec The style part is really grating. It’s like the training data through out a lot of the conventions and clarity that Ruby can provide.

[Scott Carleton @ScotterC]

@scottbelsky appreciated your answer tonight re: interest graph and preference portability. Thinking along the same lines.

[Scott Carleton @ScotterC]

@albertwenger @cursor_ai Certainly good enough to be a daily driver - and then lots of great opportunities for ongoing value-add.
I would love to do full design doc integration in cursor with the codebase. Keeping local prompt templates is a start. https://t.co/qHMPdN6lTA

[Mckay Wrigley @mckaywrigley]

Start building out a “prompts” folder in Cursor to 2x your workflow.

It allows you to create reusable pieces of frequently used instructions & context that you can give to the AI.

Totally takes things to another level.

Watch this demo of my workflow.

[Scott Carleton @ScotterC]

Improving reasoning won’t ‘fix’ hallucinations for good. There’s a whole battle beginning over influencing foundational LLMs.

How Do You Change a Chatbot’s Mind? https://t.co/rqHUwnXNzm

[Alex Rudall @alexrudall]

Love posts like this. @ScotterC https://t.co/t8fMQTIJwO #ruby #ai