An onboarding protocol for Mixer Mode (working hypothesis)

Heads up before anything else: I'm not a psychologist and I'm not a neuroscientist. What follows is a pedagogical thought experiment — an operational protocol built on my own observation of senior practitioners and on the expertise literature, not on a clinical study. Read it that way. I'm sharing it in precise form so it can be falsified.

The question I'm trying to answer: how do you train someone to hold four channels — Producto, Arquitectura, Código, QA — in parallel without collapsing them into sequential hats? The pull of the hat-era organization is strong. The default mentoring practice reproduces hat-era expertise. So if you want mixer-emergent practitioners, you have to design for it. The protocol below is my current best draft of that design.

Why 'Protocol' and Not 'Course'

Mixer Mode is expertise, in the Dreyfus & Dreyfus sense — stage 5 of the five-stage model of skill acquisition (Dreyfus & Dreyfus, 1980; later expanded in Mind Over Machine, 1986). The structural property of stage 5 is that it doesn't transmit by syllabus. The rules that constitute the skill have been internalized to the point where they no longer surface as rules, and the practitioner operates from situated judgment rather than from explicit decision procedures.

Polanyi's account of tacit knowledge in Personal Knowledge (1958) gives us the same finding from a different angle: we know more than we can tell, and the part we can't tell is the part that's load-bearing in expert practice. A course that tries to articulate Mixer Mode end-to-end and then transmit it as a curriculum will succeed at transmitting the articulable layer — which is the surface — and fail at the layer underneath, which is where the skill lives.

The closest working model we have for transmitting non-articulable expertise at scale is medical residency. Extended pairing across real cases, with the expert articulating in real time what they're attending to and why, over a period long enough that the junior's own subsidiary awareness reshapes through exposure. Russinovich and Hanselman's 2026 CACM piece argues, persuasively, that this is the model software engineering education should be moving toward in the AI era. I agree, and the protocol below is one operationalization of that intuition.

The word I'm using deliberately is scaffolding, not curriculum. A protocol provides structure for the transmission to happen — pairing cadence, articulation prompts, deliberate cases — without pretending to encode the content of what's being transmitted. The transmission remains tacit. The protocol just makes sure it has a place to happen.

Phase 1 — Structured Observational Pairing (Weeks 1–4)

The junior shadows the senior in real decisions, not in demos. The distinction matters. A demo is a curated, rehearsed walkthrough of a known outcome. A real decision is messier, has dead ends, and includes the moments where the senior corrects themselves silently. Those silent corrections are where the most important learning happens. They don't show up in demos.

After each decision, the senior performs a deliberate articulation: they walk the junior through the channels that were "in the background" during the meeting or the work session — the things they didn't say out loud but were modulating. "While we were debating the API contract, I was also tracking whether the QA timeline would absorb a change here, and whether the architectural drift this introduces would compound on the migration we have in February." Klein's Sources of Power (1998) describes this retrospective recovery of naturalistic decision-making as a teachable practice — the expert no longer notices these channels explicitly, but they can be coached to surface them after the fact.

The purpose of phase 1 isn't for the junior to learn the right decision. It's for the junior to learn that there are channels at all — that what looked like a single thread of reasoning was actually multi-channel modulation, and that the expert's apparent fluency comes from running several channels in parallel and articulating only the one that needed to be spoken. Four weeks is roughly the time required for the junior to stop being surprised by the existence of the background channels. That's the exit criterion for phase 1.

Phase 2 — Accompanied Decisions With Forced Articulation (Weeks 5–12)

The junior now takes bounded decisions, the senior present. Before executing, the junior articulates the open channels: "I'm about to commit this. The channels I see open are: API contract impact on the mobile client, test coverage on the edge case we discussed yesterday, the deploy window. I'm not seeing anything else." The articulation is the work.

The senior's job is to intervene only on the omitted channel. They don't correct the decision. They mark the missing dimension: "You're not tracking the runbook that needs updating for on-call." The junior then incorporates the channel and proceeds. The decision is still the junior's. The intervention is on the awareness, not on the outcome.

This is where Polanyi's predicted articulation difficulty shows up, often quite painfully. The junior struggles to name what they're attending to because attending and naming are different cognitive operations, and the second one is genuinely hard. That difficulty is a symptom of learning, not a defect. The senior should not interpret an inarticulate junior as an inattentive one. They should interpret it as a junior whose subsidiary awareness is developing faster than their focal-awareness vocabulary, which is the normal sequence and the desirable one.

Eight weeks at this phase isn't arbitrary. It's roughly the duration needed for the articulation muscle to start operating without conscious effort — for the junior to stop having to assemble the channel inventory deliberately every time and to start producing it spontaneously. That's the exit criterion: spontaneous, structurally complete articulation before a non-trivial decision.

Phase 3 — Deliberate Recovery as Teachable Practice (Cross-Cutting)

Phase 3 runs across all the others, because it addresses a problem the other phases create. Holding multiple channels open is cognitively expensive in a way the hat-era practice wasn't. The mixer doesn't rest by default — the load is integral across open channels, not peak on a single one. If the junior isn't taught to deliberately close channels, the load accumulates and the practitioner burns out before they reach mixer-emergent.

Csikszentmihalyi's Flow (1990) describes flow as a condition of sustainable high concentration on a single task with clear feedback and matched challenge. Mixer Mode breaks the conditions for flow — the channels don't converge on a single task, the feedback is asymmetric across channels, the challenge level varies per channel. Flow is the wrong cognitive baseline for Mixer Mode. The right baseline is something closer to a sustained, modulated multi-track operation, and the recovery model has to match that baseline rather than the flow one.

Concrete practices in the protocol: scheduled no-meeting half-days, so one channel can run at full without the interruption that lifts the others; a fader-down ritual at end of day, in which the practitioner explicitly notes the state of each open channel and the next action; a journaling practice for the channel that's consuming background attention without being acknowledged. These are not psychological interventions — I'm not qualified to prescribe those. They are operational scaffolding for the structural problem that the mixer has no natural off state. The junior learns to close channels the same way they learn to open them: deliberately, with practice, under observation.

Phase 4 — Deliberate Pressure Cases (Month 4 Onward)

The last phase introduces artificial constraint: the junior takes a decision under time pressure, with incomplete data, or with a difficult stakeholder present. The senior watches which channel the junior suppresses under pressure. Every practitioner has a suppression pattern — the channel that goes first when the load spikes — and the suppression pattern is formative information about the practitioner's developing repertoire, not a failure.

The senior names the suppression after the fact: "Under the time pressure, you stopped tracking the QA implications. That's the channel that goes first for you. Now we know." Naming it doesn't fix it. Naming it makes it visible, which is the precondition for the practitioner to work on it.

Multiple repetitions under the same type of pressure, with the suppression pattern named each time, consolidate the mixer repertoire — not by eliminating the suppression (suppression under load is structurally necessary; the mixer doesn't have infinite capacity) but by making the suppression deliberate rather than automatic. The mixer-emergent practitioner suppresses by choice, with awareness of what's been suspended and when it needs to be re-engaged. The hat-locked practitioner suppresses automatically and forgets what they suspended. The phase 4 exit criterion is deliberate suppression, not absence of suppression.

What the Protocol Does NOT Promise

The protocol has not been experimentally validated. It is a working hypothesis built on my own observation of senior practitioners I've worked with across the last few years, plus the expertise literature cited above. I have not run a controlled study. I have not measured outcomes across a cohort. What I have is structured observation and a coherent argument from adjacent fields. That's enough to publish; it's not enough to claim certainty.

It does not replace the sensitivity of a good preceptor. Senior quality is the upper bound of what the junior can learn. A protocol applied by a senior who isn't themselves operating in Mixer Mode will produce a junior who learns to articulate channels without ever learning to modulate them, which is worse than learning nothing. The senior has to be the real thing. The protocol amplifies a real preceptor; it cannot substitute for one.

It does not work in three months for everyone. Medical residency lasts years for a reason. The protocol's full four phases compress the structural elements of residency into a software engineering context, but the compression has limits. Some practitioners will reach mixer-emergent by month six. Others will be in phase 2 still at month nine. The variance is part of the model, not a defect of it.

Honest Close

I'm offering this in precise form so it can be falsified, not so it sounds correct. If you try a version of the protocol in your team and a phase doesn't transfer — if phase 2 articulation doesn't develop, if phase 4 suppression patterns don't show up, if phase 3 recovery practices don't move the burnout curve — that information is more valuable to me than confirmation. Confirmation is cheap. Disconfirmation refines the model.

The protocol is a v1. It will need revision based on field use, and that revision is the next round of work. Read it as an invitation to test, not as a manual to deploy.

If you train juniors today: which pieces of this protocol are you already doing without naming them — and which is the hardest to operationalize in your current context? Send me a DM or reach out via the contact channels at rlabs.cl.

#Engineering #FutureOfWork #Productivity #TechLeadership