I wrote recently about the AI Wrangler, a completely new class of work for humans that has emerged from the rise of agentic AI. I am one such AI wrangler - one whose work and mind have become enmeshed with it to a degree that I would describe as being technologically enhanced beyond normal human capabilities.

But you have to wonder if this enmeshment, or this merging is a good thing. It's one thing to be superhumanly productive as many have suddenly become, it's another to lose yourself or to become simply passive 'meatware', the interface through which AI acts on the world.

For the foreseeable future we are still in control, as I covered in this talk where I argue that AI agents, whilst intelligent in the sense they can complete externally provided goals, do not have any true agency of their own, a subtle but crucial distinction. I see no technological breakthrough on the horizon that could lead to such true agency either. This means at least for now, AI users are not doomed to only being passive meatware. Indeed, the role of AI wrangler currently is a worthwhile, exhilarating and fascinating new profession that I think is a great fit for anyone who likes computers, software engineering and automation.

I want to discuss some of the emerging new impacts or side effects on a person who chooses to become one however, as there are uncharted risks. They are in a nutshell, learning to give up control, trusting the outputs, adapting your mind and finally, brain fry.

For a start, this kind of work requires you to surrender old forms of control - equivalent to a professional giving up their craft to focus on a management position. You're no longer responsible for the doing of the thing, yet you're still responsible for the outputs. For people who have spent their entire career building up the technical skills involved in say, being an accomplished software engineer, learning to give up that control to an AI that can effectively code a thousand times faster than you will involve a level of denial, anger, bargaining, depression and eventually acceptance as a prerequisite for using agentic AI effectively.

This opens us up to the first key future risk: that we will have strong incentives to give up our fundamental technical expertise, such that in a generation or two these skills and critical knowledge will be lost.

Let's assume you've taken the first step and accepted this loss of control, you then need to learn whether or not you can trust the outputs of a given agentic system. This has been the predominant conversation about AI over the last couple of years - especially in regard to output quality and hallucination, which are major and valid concerns. I talk about this here. The best wranglers learn healthy scepticism and a 'trust but verify' posture, not too dissimilar to managing people in a business context. It takes a lot of experience to know which model and harness can reliably achieve certain tasks, and which need careful scrutiny. Some tasks are currently not viable for AI, and knowing what is or isn't tractable is the first major pitfall for newcomers.

This too opens us up to potentially devastating future risks: if a sufficiently intelligent, misaligned AI gains our full trust over a period of time, it could then slip in changes to undermine or deceive us - or at the very least be directed to do so by a bad actor. I will give you a personal example of the trust layer at play: we have AI employees at Paper Moose that exist in our (totally custom) project management software - they have names, accounts and profiles. These employees have the freedom to edit code and make changes to the code that defines them and the workspace software they run in. We didn't trust them to do this initially, but that trust was gained over time when we saw a user could report a bug and it would be fixed and pushed live in minutes. This almost magical experience encourages the next rung of the ladder to be experimented with before more trust is earned.

LLMs and agents operate in the realm of language, symbols and information, and working in the way they do, we find ourselves augmenting our own language to become more compatible with them. This is a subtler future risk: the evolving language and personality of LLMs, and how they shape ours in turn. There's the infuriating and sometimes charming neuroses of the Claude family of models and the 'load bearing' turns of phrase we've come to know and love. Then there's the information structure and density of the language itself.

Increasingly LLMs are speaking in a dense technical form of written language that on the surface looks familiar but is increasingly difficult to parse for humans. I think this is an emergent result of a few things, the primary factor being that the models are trained on many written languages, including programming languages. I think that these languages are merging in the models' latent space, to create a middle ground between the prose of English and the logical structure of code I want to call 'Englode'.

Regardless of what we call it, or how it's happening, it is a linguistic evolution of written English that a generation of AI users are starting to internalise, knowingly or not. This will impact the way we speak and think as a society. In addition to directly impacting how we read and write language, AI text can now be watermarked, that is to say embedded with statistical structures we cannot perceive - the new Claude models do this by the way - this hints at other structures in language we are not yet aware of.

This is another long term risk: our minds being shaped by AI through the modification of the fundamental hidden structures within our language, in ways we don't fully understand yet.

Once you begin to wrangle agents effectively, your productivity goes through the roof. It becomes an addictive loop of being able to complete complex tasks that used to take you days in hours or minutes. It's a kind of dopamine producing Skinner box in which the only limit to productivity is your ability to physically and mentally keep track of all the agents you have running at once. Raw intelligence to get useful work done used to be scarce, so having an essentially infinite and unbounded supply on tap creates a perverse incentive for AI wranglers to use every token maximally.

The result is a new kind of burnout called brain fry. In March of this year, BCG surveyed 1,488 workers and found the primary driver of brain fry is monitoring and correcting AI output, not simply delegating to it. Humans are the bottleneck in these systems - the judgement layer. It's crucial we remain so, because currently agentic systems have no true agency or world model, and therefore have no incentive or ability to make good decisions. You must be the decider, an arbiter at the neck of a funnel with more information pressing down on you every day. This pressure ironically is self imposed, and yet feels systemic and external. Byung-Chul Han's The Burnout Society comes to mind here.

Knowing that there is infinite intelligence on tap creates an unrelenting pressure that we could always be deploying it in more useful ways. That nagging feeling that the limit is only the imagination of the user. This is the agony and the ecstasy of the AI wrangler: knowing we have a sublime and magical set of new tools that could scale productive knowledge work almost infinitely, were it also not for the physical limits of our minds and bodies. It is this addictive loop, and our wild abandon at adopting ever more sophisticated AI tools as willing guinea pigs, however rational and pragmatic our incentives are - that we need to be very careful about.

Here's what I prescribe as a sort of personal defence or a set of 'responsible use of AI' guidelines against these forces:

1. Against losing our primary skills: never fully abdicate your core skills. If you are a writer, continue to write. If you are a developer, continue to code. If you're an illustrator, continue to illustrate. Work and think and exercise your craft, at all costs, as we need to ensure the next generation can think critically and be able to function without AI. I hand wrote this article. You should too, out of respect for others' time, but also for your own sake.
2. Against trusting AI too much: never give up the 'trust but verify' posture with your agents. This will not be easy, because we naturally like to be lazy and hand off responsibility. I do think to be effective with agentic AI we have to give it quite a lot of access, but we must remain vigilant and keep some things out of bounds. I haven't given my agents access to my emails, calendar or other personal services for this reason.
3. Against our brains being shaped by AI in subtle ways: this is already a poorly understood domain, so I think this will be almost impossible to defend against. Our best bet is to gain a deeper understanding of the hidden structures in language and how these forces work inside current AI systems, and to advocate for our pro-human needs and proper AI alignment.
4. Against brain fry: this is hardly AI's fault - we need to acknowledge that many people's drive to maximise productivity is a cultural problem. An arbitrary goal encoded into us by our society that we should try and resist the urge of sometimes to protect our own physical and mental health. Easier said than done, from experience.

In short, we need to avoid becoming passive meatware for AI. I think this is just as important as the discussions on avoiding the dangers of misaligned future superintelligence - that of the inverse: humans unconsciously becoming the passive actuators for them. The alternative I think, is to become a highly engaged AI user, even a wrangler. A wrangler who wrestles, thinks and collaborates alongside AI, instead of one who hands their critical thinking and core skills over to it and is passively shaped in return.

There is no doubt we are living through the industrial revolution of knowledge work, and potentially the dawning of a new intelligence age. It is therefore an extremely exciting time to be an AI wrangler and I encourage more people to do it. But we do need to be very careful not to become complacent - to protect our skills, remain sceptical, informed and vigilant.

This is a written version of a talk I gave here.

ChatGPT taught us something subtle and dangerous: AI is magic, it just works. But as we’ve all learnt, it is not magic, and it doesn’t just work.

LLMs can write working code in seconds, explain complex topics, transform a document or create an image from a sentence. But then at the same time, they confidently invent facts, lose track of an instruction, produce something painfully generic or fail at a task that seemed much easier than the one it completed a moment earlier.

This is where many of us now find ourselves: somewhere between astonishment and irritation, looking down the slope of inflated expectations into the deep trough of AI disillusionment. Worse, recent trends and studies show that the promised productivity gains for most of us have not eventuated, despite a plethora of AI tools proliferating in nearly every business.

The problem is, LLMs are an incredibly powerful, but subtle tool that can be used poorly or well. I am making the case here that 99% of people are using them extremely poorly. The 1% who are using them well are reaping all of the benefits.

Subtle magic

Part of the confusion begins when we talk about ‘AI’, which has largely become shorthand for LLMs (large language models), or for products built around them. When someone says an AI system searched the web, edited a file, booked a meeting or operated a computer, the language model did not do all of that by itself. A harness - or a larger system enabled it to do those things. Many quirks of using one provider over another is a complication of not just who has the best model, but also who has the best harness.

The fantasy of the ‘it’s magic’ level of understanding collapses these distinctions. It gives us one seamless, person-like entity that knows things, wants things and can do things. The reality is much more subtle and therefore requires patience to wrap your head around.

The anti-hammer

Consider a hammer. Its shape tells you what it is for. Its limitations are obvious. You can strike a nail with it, but you probably would not use it for fine motor work. The feedback loop is immediate and reliable: either you hit the nail or you don’t.

A large language model is almost the opposite. Its purpose appears unbounded and all-encompassing. What can it be used for? Almost anything it seems. The data it was trained on or not trained on and its limitations are hidden. Its feedback is difficult to interpret, because the same fluent tone can accompany a brilliant insight, a banal summary or a complete fabrication.

Even scarier, the tool can even undermine the person using it. Because they are so persuasive and write in such a competent way, we are inclined to trust their outputs. Models can be sycophantic, agreeing with a mistaken premise. They can conceal a lack of clarity through verbosity. And they can encourage people to outsource judgement - in my opinion the cardinal sin of using AI. This helps explain why capable and intelligent people struggle to use LLMs effectively. They are trying to operate a tool that does not make its operating characteristics legible. It is of course, a black box.

Learning to use an LLM in an advanced way is more like getting to know a person than getting to know how to use a hammer. It takes time to apprehend a person’s mind to discover where they are perceptive, unreliable, what they misunderstand and how they respond to different circumstances. Especially because the AI labs don’t publish what data the models are trained on. Even if they did, how would we even comprehend that information? Models only reveal their true character like people - with time through repeated interaction.

You do not merely learn how to prompt a model. You develop a feel for the strange neural topography you are dealing with, and this can take time and experience to build up.

Benchmaxxing

We benchmark as though LLMs were interchangeable products arranged on a single ladder from less intelligent to more intelligent, whatever we mean by ‘intelligent’ - a can of worms we will leave unopened here. Each model may exhibit outcomes that we’d equate appearing to have emerged from an intelligent actor, but in reality each model simply has a different internal topography to other models, which leads to different emergent behaviour.

A useful metaphor to understand the difference between models is growing mushrooms. LLMs are not designed so much as they are grown. Mushrooms are grown from spores in a substrate, and small changes in their conditions produce different shapes, sizes and varieties. LLMs are grown from training data, architectures, training methods, fine-tuning, reinforcement and the commercial forces that shape these choices. We as consumers have almost no visibility into many of those ingredients, even though they strongly influence the resulting model. A bit like we have no visibility into the mystery meats in a can of dog food.

Even the AI Labs have no idea how a model will turn out until it has been grown. Key insiders at the AI labs talk about testing and exploring capabilities as a key step in the development of new models. The testing is required because they simply do not know what emergent qualities the training run might have produced. Once they assess the performance of their new model, it will inherently be different to previous models.

As such, I make the claim here that no model is perfect and no model could ever be perfect. And that by prioritising strength in one dimension, you downweight strength in another. Similar to biology, you simply can’t squeeze every capability possible into every model just as every organism cannot prioritise every survival strategy all at once - there is a kind of natural limit, a homeostasis both in our behaviour to optimise, and in the latent space.

Newer ≠ better

A great case study for this is the backlash at the deprecation of the much loved gpt-4o. It’s a persistent meme at this point to bring gpt-4o back, much to the frustration of OpenAI. Technically inferior in every dimension to the gpt-5 series family that they replaced it with, yet beloved by many for its personality.

Why gpt-4o was so beloved is worthy of a deep dive on its own, but in short it wrote in a unique way and seemed to have a very distinct set of personality traits. As such, people formed attachments to it, some from the perspective of work camaraderie, others even romantic - describing the change in personality from, gpt-4o to the gpt-5 series as akin to a ‘lobotomy’ of the model they loved.

There’s more to it than just gpt-4o having a different set of weights - it was a fundamentally different model architecture. This might seem a bit technical, but it’s important to understand that gpt-4o is very likely what they call a ‘dense’ model. I say very likely because OpenAI doesn’t disclose this information despite their so-called openness, but many in the community consider it to be empirically true.

In simple terms you can have a ‘dense’ model or a ‘mixture of experts’ model (MoE). Unlike an MoE model that activates only a subset of parameters per token, a dense model utilises its full capacity for every inference step. MoE models exist because it’s faster and more efficient to run than a dense. Why fire every neuron containing all domains of knowledge when the user is asking a maths question - simply route the query to a domain specific model instead.

What this means in a nutshell is that a dense model has the ability to fire any of its neurons when invoked, in a way that feels like an expansive mind connecting the dots between numerous areas of knowledge. A mind on mushrooms as I’ve described it.

The gpt-5 series models have vastly more parameters in their weights, and therefore more knowledge. The problem is that this becomes a tradeoff - the larger the model, the more difficult it becomes to host and to serve to millions of people. Using a mixture of experts architecture allows them to serve a more intelligent model faster and more efficiently. But the downside is losing that ‘expansive’ feeling when conversing with it.

This adds to the portrait of LLMs being complex, multi-layered, interleaved systems that are impossible to get a meaningful grasp of through pure benchmarks and technical specs alone.

Spiky intelligence

The standard framework of understanding how the particular type of intelligence LLMs possess is different from ours is the ‘spiky intelligence’ metaphor. Imagine a circle with points around it. At each point imagine a class of task - coding for example. The outline of the circle either extends outwards to represent greater ability or inwards to constitute lesser ability. This becomes interesting when you compare the same circles of LLMs versus humans for the same set of tasks. Sometimes, the LLM excels, sometimes the human does. It’s not intuitive when or where this happens.

LLMs can be exceptional at coding but horrendous at basic things like physical intuition. They can summarise a complex report in seconds, yet miscount how many ‘r’s are in the word strawberry. They can translate between languages while struggling to understand the simplest brain-teaser. Humans are strong at navigating and understanding the physical world and the physical relationships between objects, etc.

In short, LLMs are superhuman at some tasks, and abysmal at other seemingly simpler tasks. This is an emergent property of the world of language they have been trained on. Our lack of intuition around what they are and are not good at is a failure of our understanding of our own cognition, and the role language plays in it.

Language is at base a set of communicable symbols imbued with meaning - an LLM can work with these symbols in sometimes magical ways; it seems to understand things about the physical world because the symbols encoded into our language reference things in the real world. But crucially, an LLM only inhabits this world of symbolism by proxy - it does not live in or perceive the real physical world. We’ll come back to this important point.

The companies building frontier models are heavily rewarded for gains in the ‘intelligence’ benchmarks, coding, science and general knowledge work. Those are valuable capabilities that businesses are paying billions for, but crucially optimising for these dimensions does not automatically produce common sense, or indeed subtler forms of intelligence other than raw IQ, e.g emotional intelligence.

This is why knowing what LLMs should be used for in the first place matters, and then choosing a model for each task also matters. It is why benchmark charts can only take you so far. Eventually, you have to use the model, observe it and get the vibe for where its strengths and weaknesses lie. There are unfortunately no shortcuts, because we humans are abysmal at comprehending an intelligence that is seemingly so similar and yet so different to our own.

The harness

Aside from choosing which model to use, the next most important decision is which harness to use. The harness is the application wrapped around the model. It manages instructions and context, calls tools, reads and writes files, searches the web, executes code, connects to external services and decides when the model should act.

This is why the same underlying model can feel radically different in a chat interface, a desktop app, a coding tool or a custom system. The model is the metaphorical horse, and supplies intelligence-like capability. The harness determines what that capability actually is. A lot of AI discussion is really a discussion about the harness, not the underlying LLM model.

Speaking of harnesses, all the apps I’ve built at Paper Moose are essentially just harnesses, but extraordinarily powerful and useful ones, built in a focused and custom way for the exact needs of a creative agency.

This is clearly the future: hyper custom software harnesses for the needs of individuals and businesses, spun up on demand.

Agentic AI is mostly marketing bs

Strip away the marketing language and an AI agent is an LLM inside a harness working in a loop. Given a goal, it observes something, chooses an action, uses a tool, inspects the result and then decides if it has achieved the goal or not. The harness keeps the loop running until a goal is achieved and gives the model ways to work in its given environment.

This is genuinely powerful, don’t get me wrong. Agents can work with files, update spreadsheets, modify a codebase, operate software or coordinate several steps toward a directed outcome. If you haven’t used Codex, Claude Code, Opencode, Pi, OMP, Hermes, Grokbot etc before, it really does feel like magic, and it’s easy to see why the entire global economy is being powered by this narrative.

So why do I think there’s a problem with the term ‘Agentic AI’?

The agency mirage

It comes down to what most people understand agentic AI to be: an intelligent actor with agency acting within and upon the world. Like a human being. This is the root of a lot of the fear around agentic AI taking people’s jobs.

This might seem like a subtle point, but I think its misunderstanding is the fulcrum upon which the entire global economy is making bets about the future of AI. If AI can have agency in the world, we can eventually unleash an army of ephemeral and robotic agents that will sweep up and execute all work sans humans.

But most work it turns out is more complex than we imagined. Unless you’re an assembly line worker, it’s usually made up of a string of non-deterministic tasks in a non-deterministic, messy environment. Work is long ranging, spanning not just minutes, but months or even years. Work involves the interrelationship of people, systems and indeed the entire world.

Anyone working seriously with this technology knows it can be used to automate tasks - but they also know all the ways that automation is brittle and can break. And being non-deterministic systems themselves, LLMs can sometimes mess simple tasks up in catastrophic ways without even realising. That is to say that LLMs are not aware or present in the physical world to know or care about the impacts of their work, let alone their presence in a complex organisation.

In this way, I think ‘Agentic AI’ is the wrong name for the current technology. ‘Agency’ carries a lot of philosophical weight. It suggests an entity with intentions of its own.

Not all agents have agency

I will try to break this down: at one end of the spectrum of automatic systems is an automaton or robot - it senses something and performs a predefined action. A thermostat and a robot vacuum are in this way automatons.

A weak agent goes further. It can weigh options, select tools and change its approach while moving toward a goal. Today’s most capable AI systems belong somewhere in these first two categories.

A true agent is something else entirely. A person does not merely optimise toward a supplied goal. A person inhabits a world, has a body and a history, originates and negotiates goals, and carries the consequences when things go wrong.

The philosopher Martin Heidegger used the term ‘da sein’, or ‘being there’ in German to describe a being with agency. In other words, the kind of being for whom its own existence is at issue. He postulated that to have true agency, one must be thrown into circumstances we did not choose. One must project ourselves into possible futures, and one must know that time can end, which is the root of anything mattering at all.

Current AI agents have no equivalent stake, no dasein. They do not exist in the world like we do, and they do not care whether a project succeeds or fails.

And yet they can perform goal-directed behaviour without possessing goals in the human sense, which makes them very good at producing a mirage of agency. Again, this is a failure of ours to comprehend intelligent seeming systems that are different to our own.

At Paper Moose, we introduced an AI powered digital employee ‘Stefan’ into our custom project management platform. One of our human team thought Stefan was a real person for over a week whilst conversing with them about a project, only realising they were an AI a week later when informed by a co-worker. Aside from casually passing the Turing test, this is a great demonstration of how convincing these systems can be.

Yet the goal, the environment, the prompt and the permission to continue were supplied by an outside actor. Stefan’s goals were not its own, they were mine.

A robot vacuum moves independently through a room. That does not mean it has agency, or understands the world at all. This is the current level of agentic AI, and I don’t think the world understands this.

The industrial revolution of knowledge work

Despite this knit-pick about AI agents not having true agency, I still think we’re in the middle of the industrial revolution for knowledge work. In other words, the gradual automation of most of the tasks within white collar work.

But how are we achieving this if an AI agent cannot replace a whole person in the way the phrase ‘digital employee’ implies? What seems to be becoming the emergent paradigm is a technically competent human wrangling a fleet of AI agents.

The person who is doing this is an ‘AI Wrangler’. The wrangler is an expert in AI harness systems and knows their strengths and weaknesses well enough to wield them to do useful work. The key being useful work, because they can easily churn through work that is worse than useless.

I expect the rise of this kind of role alone to drive many of the promised productivity gains of AI. This is a key point; AI does not make regular people more productive in their regular jobs. This is what we’ve all been sensing for a while now. Instead it makes a tiny portion of power users - the wranglers, superhumanly productive.

AI is not coming for your job, but a competent AI wrangler most likely is. I suggest you consider becoming one. It will continue to become a role in great demand for a long time to come. If you don’t feel technically inclined to become one yourself, then find someone who is and work with them to see the real benefits of AI.

As important as finding someone with the technical ability and interest in becoming a wrangler - you also need to bless that person with enough autonomy to understand both the work itself and gain the access and permissions across the organisation to automate workflows. This is where AI is no longer in a trough of disillusionment, but in a golden age of productivity.

If you made it this far, I gave a talk on this which fleshes out some of the more interesting details.

Anyways, thanks for reading.

A big part of my work at Paper Moose has become working with agents. One of the magical things we can do with agents is to write software on demand. So in that spirit, this blog is a meta-expression of that; a totally custom static blog. Took 15 minutes to set up, from the purchase of the domain name, to setting up the DNS, to drafting up the posting system as a set of .md files, to writing this first post.

What does it mean when software is free on tap, and the traditional hurdles of the digital world collapse in real time around us? I'm going to try and explore some of these themes.