Industries where the framework does not reach
Disclaimer up front, and this is the last one of the series: I'm coming off seven posts translating Mixer Mode into industries that aren't mine. Before closing, an honest note. There are domains where the frame doesn't reach, and if I don't name them, the frame becomes hype.
A framework that claims to explain everything explains nothing. So this post is the discount on the previous seven — the places where I'd ask you to put the lens down.
Why This Post Exists
A framework with no scope conditions has no edges, and a thing with no edges isn't a tool — it's a worldview. I'd rather have a sharp tool with a clear handle than a worldview that survives every counter-example by quietly re-describing it.
The series translated Mixer Mode into six adjacent industries: government, finance back-office, ERP consulting, healthcare administration, legal practice, education. All of them are cognitive work over information. All of them involve a senior practitioner holding multiple concerns at once over a deliverable that is, ultimately, readable.
But there are three categories of work where the lens breaks. Naming them is part of structural honesty, and it's also the only way the framework continues to be useful in the places where it does apply. So here they are.
Physical Domains (Surgery, Welding, Trades of the Body)
A surgeon runs multiple simultaneous concerns — yes. So does a welder doing a critical structural joint, so does an electrician working live wiring, so does a mechanic diagnosing an intermittent fault.
But the binding constraint in those domains was never cognitive in the sense Mixer Mode describes. It was motor skill, calibrated under pressure, in the presence of living tissue or molten metal or working voltage. The judgment is real and the simultaneity is real, but the execution layer — the hands, the tools, the body — is the part that's actually scarce.
The agent doesn't lower the cost of execution for the surgeon the way it lowers it for the programmer. The cognitive scaffolding around the procedure can be aided: pre-op planning, intra-op decision support, post-op tracking. The procedure itself is still hands on tissue, and that part doesn't get cheaper because an LLM exists.
Mixer Mode can apply to surgical planning. It doesn't apply to surgical execution. That boundary is worth being explicit about.
Presence Domains (Therapy, Relational Sales, In-Person Teaching)
There's a second category that's harder to talk about, because the work looks informational from the outside but isn't, structurally.
A therapist holds channels — yes. So does a teacher with a difficult class, so does a sales lead managing an enterprise relationship across a multi-quarter cycle. The simultaneity is there.
But the value being produced is not information. The value is in the human bond — the fact that another person experiences being met, being attended to, being responded to by someone present in the room with them. The setpoint is the person across from you, not a producible system that can be observed by Meta-Software.
An LLM can simulate the output of that work. It can transcribe the session, summarize the themes, draft a plan, write a follow-up. What it can't do is hold the presence. And the presence is the thing the client is actually paying for, whether they would describe it that way or not.
Mixer Mode names cognitive simultaneity over information. It does not name what happens when another human being needs you to be there. Those are different categories of work, and I think the framework loses its precision if I pretend they're the same.
Domains of Body or Matter (Crafts, Tangible Art)
The third category is the one I'm most cautious about, because it's the one where my own instincts run weakest.
The carpenter, the ceramicist, the cook, the sculptor. Their knowledge is tacit in the body, in Polanyi's sense — and arguably tacit-squared, because not only is the practitioner unable to fully articulate what they know, the apprenticeship doesn't transmit it through articulation. It transmits through years in the workshop, under the eye of a master, doing the work badly until the body starts to do it well.
Meta-Software requires the output to be inspectable by software. A line of code is inspectable. A pot pulled on a wheel, evaluated for the way it sits in the hand, is not — at least not in any operationally useful sense for the kind of supervisory layer the framework describes.
The framework talks about cognition over information. It is not a framework about all work. The craft domains are work, and they are skilled, and they will be affected by AI in real ways. But the affecting will look different — it won't be the Mixer Mode / Meta-Software pattern. It will be something else, and I don't have a name for it.
What Is Clear at the End of the Series
Mixer Mode is portable to cognitive work over information — and within that domain it travels fairly well, as the seven previous posts have tried to demonstrate. The lens picks up real structure in software engineering, ERP consulting, financial back-office, public administration, legal practice, healthcare administration, and education.
Meta-Software requires that the agentic output be inspectable by software. That's a non-trivial condition. It rules out the presence work and the body work, and it limits the framework's reach to the deliverable-as-text domains.
The domains where it applies are large but bounded. Saying that out loud — drawing the perimeter — is what separates a usable frame from a metatheory that explains nothing.
A Note on Scope Conditions
The original paper phrases it more carefully than I just did: the framework "bites fully now in organizations whose primary product is software". I'd extend that, conservatively, to: it bites in any domain where the deliverable can be read.
Where the deliverable is presence, the lens fogs. Where the deliverable is body, the lens fogs. Where the deliverable is matter, the lens fogs. In those places I'd rather hand the conversation to people who actually work in those domains than insist the framework reaches further than it does.
That's the close. The framework is useful in the places it's useful; it's also bounded, and the bounds are real.
This closes the series. If it helped to see the lens applied to your industry — or to see where it doesn't reach — DM me or reach out via rlabs.cl. And if you work in one of the "non-touching" industries and think the lens does apply somehow, that debate interests me more than the agreements.
#MixerMode #FutureOfWork #DigitalTransformation #AI