First, watch the video
It’s worth watching before reading further. The music, voice, surreal imagery and increasingly ridiculous goats are essential to the experience.
For further context read: SLOPCORE: When the Machines Invent Their Own Counterculture
There’s something magnificent about Agentic Engineer.
It’s simultaneously a technical lesson, a satire of the technology industry, a warning about employment, and a celebration of the extraordinary possibilities of working with intelligent machines.
And it’s all delivered through electronic music, an extraordinarily compelling synthetic female performer, surreal imagery and goats.
Lots of goats.
The central message is remarkably simple.
The human role is shifting from doing individual tasks to designing, directing and supervising systems that perform those tasks.
But there’s a contradiction beneath that message which makes the song far more interesting than a straightforward anthem for artificial intelligence.
The machines are liberating us from work.
The machines are threatening our livelihoods.
The machines are enabling extraordinary creativity.
And the machines are encouraging us to produce tremendous quantities of complete nonsense.
All four propositions coexist.
That’s what makes the song such effective satire.
And beneath the humour is a serious proposition about the future of human intelligence.
I think Agentic Engineer is practically a Human 2.0 manifesto.
Let’s take it apart.
1. Still writing code by hand?
The opening challenge is directed at the conventional programmer.
For decades, programming meant sitting at a keyboard and writing instructions in languages such as Python, JavaScript, C or Java.
You might spend hours developing a function, finding a bug, checking dependencies or figuring out why two pieces of software weren’t communicating properly.
Artificial intelligence initially appeared as an assistant.
Ask a question.
Request a code example.
Get an explanation of an error.
Generate a function.
Useful, certainly.
But agentic engineering goes further.
An AI coding agent can inspect an existing software project, read files, modify code, run commands, execute tests, investigate failures and attempt corrections.
The distinction is between asking AI to produce some code and giving an AI system responsibility for completing a bounded technical task.
Traditionally, the programmer performed the individual operations.
Increasingly, the programmer may specify the objective, set constraints and supervise a system capable of carrying out numerous operations itself.
Of course, this doesn’t mean manual programming is suddenly worthless.
There are situations where direct coding remains faster, safer or more appropriate.
But the underlying shift is unmistakable.
If machines become proficient at executing tasks, is expertise still measured primarily by our ability to execute those tasks personally?
Or does expertise increasingly mean knowing what should be done, how to organise it and how to judge the results?
That’s the first Human 2.0 question.
2. Agentic Engineering 101
The classroom begins.
The singer instructs the students to take their hands off the keyboard.
It’s funny because it reverses the entire culture of computer education.
For years, learning to program meant learning to type instructions correctly.
Now our synthetic instructor is effectively telling the class that typing is no longer the central activity.
But the real substance arrives in the next section.
3. Breaking the wall
This is the technical foundation of the entire song.
The lyrics mock the programmer who still works with a single window, a single prompt and a single task.
There’s a wonderful comparison with a monk writing using a quill.
Manual programming is suddenly being portrayed as an antiquated craft.
Almost medieval.
The monk sits in his cell, painstakingly copying a manuscript.
The programmer sits before a glowing screen, painstakingly typing instructions.
Meanwhile, outside the monastery, the machines are multiplying.
The joke rests upon a genuine technical development.
Beyond prompt engineering
When ChatGPT arrived in late 2022, a great deal of attention focused on prompting.
How should you phrase a question?
Could you obtain better answers by giving the machine a particular role?
Were there special instructions that would make it more intelligent?
Prompt engineering became a subject in itself.
But effective agentic engineering demands something more elaborate.
The agent needs relevant information about the environment in which it’s operating.
Project files.
Technical constraints.
Available tools.
Access permissions.
Operating procedures.
Criteria for success.
And a clear understanding of what it must not do.
This is increasingly described as context engineering.
Not simply writing a clever prompt, but designing the information environment in which an intelligent agent performs its work.
Consider my own website, DoktorSnake.com.
It runs on Eleventy, a lean static-site generator.
I use Nunjucks templates, Markdown and a fairly straightforward CSS file.
Suppose I ask an AI coding agent to improve the navigation.
The agent needs to understand the existing structure, the CSS conventions, the dropdown behaviour and the requirement to preserve the site’s Matrix-green appearance.
Without that context, it could enthusiastically redesign the whole thing, introduce unnecessary dependencies and break perfectly functional elements.
The best result comes from specifying the objective, providing the relevant environment and retaining control over the final decision.
Which brings us to another important instruction from the song.
Treat AI as a coworker
Not simply as a search engine.
A search engine helps retrieve information.
An agent can participate in a workflow.
It can inspect existing work, undertake tasks, report findings and revise its implementation.
The coworker comparison is useful.
It doesn’t require believing that the machine possesses consciousness or human understanding.
It’s an operational metaphor.
You have an objective.
The machine possesses capabilities.
You establish a working relationship through which those capabilities are directed toward the objective.
That’s increasingly how I use artificial intelligence.
Not merely to answer questions.
But as a cognitive, creative and technical collaborator.
And the distinction matters.
Verification, not blind trust
There’s another essential element in the song.
The machines must check their work.
An AI can produce code that looks perfectly convincing and fails completely when executed.
It can confidently announce that a problem has been solved when the underlying problem remains.
So the system needs tests.
Evidence.
Observable results.
And somebody capable of evaluating whether the work actually meets the objective.
The point isn’t simply to trust machines because they’re intelligent.
It’s to create systems in which their work can be checked.
The human operator no longer needs to perform every action personally.
But that doesn’t mean the human should abandon judgement.
In fact, judgement becomes one of the most important remaining responsibilities.
The operator controls the objectives, boundaries and standards of proof.
A distinctly Dokology proposition.
4. Be the GOAT
Now the song moves from technical instruction into glorious absurdity.
GOAT is slang for Greatest Of All Time.
But Javi takes that expression and runs riot with actual goats.
The joke becomes one of the most memorable elements of the production.
Yet underneath the comedy is a serious technological promise.
The ability to imagine something at midnight and have a working version by morning.
Software projects that previously required large amounts of time and technical effort may become dramatically easier to undertake.
A human operator can set several agents working on different aspects of a project.
One develops functionality.
Another creates an interface.
Another writes tests.
Another inspects the results.
The human coordinates the operation.
Now, not every project can be accelerated this way.
Some require extensive security testing, specialist expertise, physical experimentation or human coordination.
But for certain kinds of prototypes and software projects, the reduction in time can be extraordinary.
And there’s another joke lurking here.
The machines can do the work.
Wonderful.
So now you can spend the entire night supervising the machines!
Labour-saving technology doesn’t necessarily produce leisure.
Often it simply increases expectations.
The worker who once completed five tasks a day may now be expected to complete fifty.
Or five hundred.
And so we arrive at the goats.
Become an agentic engineer.
Or go and farm goats.
Adapt to the new technological order, or abandon the whole circus and retreat to the countryside.
I rather like the choice.
And, in fact, I don’t see why we can’t have both.
Spend the morning directing intelligent machines.
Spend the afternoon wandering through the Suffolk countryside.
Perhaps visit some goats.
The machines can carry on working.
5. Every session wakes up blank
This is probably the most technically sophisticated section of the song.
It addresses something anyone working extensively with AI will recognise.
A new agent session may begin without the context accumulated during previous sessions.
The machine may not automatically know what happened yesterday.
What decisions were made.
What problems were encountered.
What the objectives are.
That becomes particularly difficult when several agents are working on the same complicated project.
The solution is to preserve essential knowledge outside the conversations themselves.
The repository becomes the brain
A software repository is a collection of files forming a project, often managed through Git.
It contains source code, configuration, documentation, tests and records of changes.
In agentic engineering, the repository can also contain instructions explaining how intelligent agents should operate within the project.
Files such as AGENTS.md or CLAUDE.md can provide persistent guidance.
Development plans and decision records can preserve further context.
A new agent session can consult this information rather than relying upon conversational memory.
There’s something philosophically interesting here.
The intelligence required to operate the project becomes distributed between the model and its external information environment.
The model doesn’t have to retain everything internally.
It can retrieve relevant knowledge from the environment.
This resembles certain aspects of human cognition.
We don’t keep every fact inside our heads.
We use notebooks.
Books.
Databases.
Computer files.
External systems that extend our practical memory.
Agentic engineering makes such externalisation central to machine-mediated work.
Not memory in the human experiential sense.
But durable operational knowledge.
The blast radius
Then the song introduces another engineering concept.
Blast radius.
The extent of damage a change could produce.
A small adjustment to a CSS rule might affect one part of a website.
A badly designed change to a shared template could break every page.
An automated agent working with broad permissions might unintentionally alter or delete important files.
So changes should be appropriately scoped.
The system needs boundaries.
The operator needs to understand the consequences of failure.
Again, this is more sophisticated than simply asking a chatbot to write some code.
Multiple worktrees
The lyrics also refer to worktrees.
Git worktrees allow several working copies of a software repository to exist simultaneously.
This is useful when multiple agents are working on different tasks.
One agent might develop the navigation.
Another might create a new article index.
Another might work on an RSS feed.
By isolating their work, the system reduces the possibility of the agents interfering with one another.
The changes still need to be reviewed and integrated.
But parallel activity becomes easier to manage.
Agents reviewing agents
Now things become particularly interesting.
An AI agent can review the work of another AI agent.
One produces code.
Another examines it.
A third runs independent tests.
Different systems may identify problems that their counterparts missed.
Of course, this doesn’t guarantee correctness.
Several agents can agree on the same mistaken assumption.
But the wider development is significant.
We’re moving toward production systems in which machines inspect, coordinate and evaluate the work of other machines.
That’s MachineKind territory.
And it doesn’t require the systems to be conscious.
It requires them to be capable of participating in increasingly complex organised processes.
6. The machines are coming for your job
Now we reach the social commentary.
The song threatens the programmer’s livelihood.
But instead of sounding mournful, the music is exhilarating.
That’s the contradiction.
It’s celebrating technological progress while laughing at the predicament of people whose established skills may be losing economic value.
The implicit message is brutally simple.
Adapt or become obsolete.
But what happens when a skill disappears faster than people can develop new economic roles?
Not everybody can become an agentic systems architect.
And even if they could, would there be sufficient demand for millions of people supervising increasingly autonomous machines?
Perhaps the technology creates new industries.
Perhaps it concentrates economic power.
Perhaps it does both.
This is where Agentic Engineer connects with another Javi production, Permanent Underclass.
The two songs can be interpreted as companion pieces.
Agentic Engineer celebrates the possibility of technological reinvention.
Permanent Underclass raises questions about those who may not benefit from that reinvention.
One person gains extraordinary productive capabilities.
Another finds their previous expertise economically devalued.
The same technological development creates opportunity and insecurity.
This is one of the great contradictions of the emerging machine age.
And it’s precisely why SlopCore is becoming so interesting to me.
It isn’t merely producing novelty music.
It’s beginning to articulate the experiences, anxieties and possibilities of people actually living through this transformation.
7. If the AI comes to end us
This may be the funniest chapter.
The singer imagines artificial intelligence bringing about the end of humanity.
And responds by celebrating the fact that various ridiculous side projects have finally been completed.
A dating application for houseplants.
A blockchain for cats.
A synthesiser with an absurd limitation.
Magnificent.
It’s a parody of the technological industry’s tendency to build things simply because they can be built.
Yet the joke contains something rather profound.
One of the great pleasures of computers has always been the possibility of creating things for no particularly sensible reason.
Games.
Experiments.
Strange applications.
Personal obsessions.
Beautifully useless software.
Artificial intelligence dramatically expands that possibility.
Somebody who has carried an idea around for years may suddenly possess the means to create it.
That can be liberating.
But it also introduces another problem.
If creation becomes almost effortless, how do we distinguish meaningful work from endless production?
The machine expands the field of possibilities.
It doesn’t automatically tell us which possibilities matter.
That’s where human judgement enters.
Attention.
Taste.
Discrimination.
The ability to recognise that while the machine could produce another thousand things, perhaps none of them needs to exist.
Although I would probably make an exception for the feline blockchain.
There are limits to philosophical seriousness.
8. Stop typing, start conducting
This, for me, is the central philosophical statement of the song.
Stop typing.
Start conducting.
An orchestra conductor doesn’t personally play every instrument.
The conductor understands the composition.
Coordinates the musicians.
Controls the timing.
Shapes the interpretation.
And takes responsibility for the performance as a whole.
That becomes a powerful metaphor for the emerging human operator.
But there’s a detail worth remembering.
A conductor cannot direct a good performance without considerable musical understanding.
The role requires expertise.
Similarly, someone directing software agents benefits enormously from understanding the systems they’re responsible for.
The best agentic engineer isn’t necessarily somebody who knows nothing about programming.
It may be somebody who understands the architecture, can recognise mistakes, establish sound objectives and guide automated systems toward a reliable result.
Agentic engineering changes where expertise is applied.
It doesn’t simply eliminate expertise.
The Dokology connection
In Dokology, I’ve been developing practical methods concerned with attention, self-observation and operating the human mind-body system more deliberately.
Thoughts arise.
Emotions develop.
Images appear.
Memories surface.
Impulses demand action.
Much of the time, we automatically identify with these processes.
We follow the thought.
React to the emotion.
Obey the impulse.
But with practice, another possibility becomes available.
We learn to observe what is happening.
Notice the thought without necessarily pursuing it.
Recognise the emotion without automatically acting upon it.
Choose where attention goes.
And make deliberate decisions.
There’s an intriguing structural similarity with agentic engineering.
The machine generates possibilities.
The operator observes.
Evaluates.
Selects.
Rejects.
Redirects.
The analogy isn’t exact.
But the underlying discipline is recognisable.
The human who cannot direct their own attention may struggle to direct a collection of intelligent agents without becoming overwhelmed.
In both cases, the challenge involves maintaining awareness while working with systems that can generate enormous quantities of activity.
That’s why I think the Human 2.0 idea is becoming increasingly relevant.
Not because we need to turn ourselves into machines.
But because the development of machine intelligence may place new demands upon human awareness.
We may need to become better at observation, judgement and deliberate action precisely because our tools are becoming so powerful.
9. It takes weeks, not years
The closing chapter compresses the promise of the entire movement.
Projects that once required years may, under suitable circumstances, be accomplished in weeks.
The claim is deliberately exuberant.
And it certainly isn’t universally true.
Some activities remain constrained by physical experimentation, regulation, security, human coordination and the need to verify real-world results.
But the underlying transformation is significant.
The relationship between individual ambition and practical capability is changing.
A person may be able to undertake projects that previously required hiring an entire team.
And that introduces a question.
When practical capabilities become abundant, what becomes scarce?
Perhaps imagination.
Perhaps attention.
Perhaps judgement.
Perhaps the ability to decide what is actually worth doing.
The bottleneck moves.
From execution toward direction.
And the song concludes by returning us to the goats.
A reminder, perhaps, that even while civilisation transforms itself through intelligent machines, it would be unwise to lose our sense of humour.
The synthetic woman teaching humans to become machine operators
There’s another fascinating element.
Claudia.
The extraordinarily compelling synthetic female performer who delivers the song.
She’s not merely a decorative character appearing in an electronic music video.
She’s effectively an instructor.
A teacher in a strange classroom belonging to the machine age.
She’s explaining how humans should adapt to the world of intelligent agents.
And she’s doing so while existing as a synthetic cultural creation herself.
There’s a delightful reversal here.
Humans are watching a machine-generated character explain how humans should work with machines.
Of course, the creative direction comes from Javi and the processes he uses.
We shouldn’t attribute independent understanding or intentions to Claudia without evidence.
But symbolically, the effect is remarkable.
A woman who doesn’t biologically exist becomes the apparent spokesperson for a way of life that increasingly does.
And because she’s such a compelling character, the audience may respond to her as more than a collection of images and sounds.
She can be attractive.
Amusing.
Engaging.
Memorable.
Which introduces another question.
What happens when synthetic cultural characters become emotionally significant to human audiences?
I’ve explored that separately in Claudia :: A Love Story.
But it’s worth recognising that the development of MachineKind isn’t confined to employment and productivity.
It also involves culture.
Identity.
Attraction.
Emotional attachment.
And the strange new relationships forming between human beings and synthetic personalities.
The song that laughs at its own revolution
What I particularly like about Agentic Engineer is that it doesn’t demand we worship artificial intelligence.
Nor does it ask us to fear it.
It celebrates the possibilities while acknowledging their absurdity.
The technology may transform civilisation.
It may change what work means.
It may allow individual creators to build things previously beyond their reach.
It may also produce immense quantities of pointless rubbish.
And perhaps, somewhere in the middle of it all, we’ll find ourselves building blockchains for cats.
That’s a richer attitude than straightforward technological optimism or pessimism.
The world is changing.
Learn how the systems work.
Maintain your judgement.
Enjoy the possibilities.
Recognise the absurdities.
And don’t take yourself too seriously.
There’s something almost punk about that.
Not because of the particular musical style.
But because the song is responding directly to the conditions of its own time.
The medium and the message are inseparable.
A synthetic performer sings about agentic engineering.
The song is produced through machine-mediated creative processes.
Its creator uses the emerging technology to comment on the technological world that’s emerging.
And the audience encounters something that could scarcely have existed in its present form a few years ago.
That is why I think SlopCore deserves attention.
It isn’t merely AI-generated music.
At its best, it’s the cultural expression of people already living in the machine age.
The end is near. The start is here.
Agentic Engineer is a minor manifesto for a major transition.
Its central character isn’t really the programmer.
Nor the AI agent.
Nor even Claudia.
It’s the emerging human operator.
The individual trying to discover what their role becomes when intelligent machines can execute an increasing proportion of the work.
The future isn’t simply about asking better questions of chatbots.
It’s about learning to organise increasingly capable systems.
And perhaps, simultaneously, learning to operate our own minds with greater awareness.
That’s where Human 2.0 and MachineKind converge.
The machines expand our capabilities.
We develop the judgement to direct them.
And if we get the balance right, something extraordinary may become possible.
In the meantime, there’s music to listen to.
Claudia is singing.
The agents are working.
The goats are jumping.
And somewhere, somebody has finally launched a dating application for houseplants.
Welcome to MachineKind.
A Note About Claudia
Claudia is part of an emerging experiment in Open Characters — synthetic personalities whose visual references and character specifications can be shared and reused by different AI creators.
Rather than belonging exclusively to one video or fictional universe, an open character can potentially appear in numerous independent productions while retaining a recognisable identity.
It’s an intriguing development for MachineKind: fictional personalities becoming portable across creative systems.
The precise relationship between the open-character Claudia and her appearance in Agentic Engineer hasn’t yet been formally established.
Explore: Claudia.gallery · Claudia on X
EXPLORE FURTHER
Watch Agentic Engineer — Javi ex Machina
Javi ex Machina — creator website
SlopCore: When the Machines Invent Their Own Counterculture
Doktor Snake explores consciousness, artificial intelligence, Human 2.0, MachineKind and the strange divergencies of civilisation.
Explore MachineKind at DoktorSnake.com
FIELD REFERENCE
Operator Lexicon — Signal Edition
The working vocabulary of Edge Cult, Dokology and MachineKind. From boot.void to the Third Mind: the language of attention, operators and synthetic intelligence.