A manager of infinite minds

Planted 02026-01-01

The constraint is orientation.

The strategy memo looked perfect. Well-structured, professionally argued, footnoted. It took four minutes to generate. It took six weeks to unwind the decisions made from its confident, subtly wrong framing.

The collaboration didn’t fail. It happened exactly as asked. The failure was upstream—in what wasn’t said, in context that seemed too obvious to state, in judgment that had never been articulated because it had never needed to be.

Most people skip straight to output. “Write this memo.” “Build this feature.” “Draft this strategy.” They get something back. It looks professional. But it misses in ways that are hard to articulate and expensive to discover.

Orientation is the hard part—the context, standards, and intent that shape whether an output is useful. Without orientation, you get production without judgment.

Output needs orientation. The work has always been orientation.

The constraint

We no longer have a shortage of minds. I can spin up a research agent when I wake up, have code written while I’m in the shower, delegate a first draft to something that never gets tired. The constraint is no longer cognitive labor. The constraint is direction.

The constraint is everything I know but haven’t shared. The context I’ve absorbed over years that tells me which opportunities are real and which are mirages. The pattern-matching that fires before I can explain why. The thing that makes me pause when I can’t articulate the problem.

Michael Polanyi gave this a name: tacit knowledge. We know more than we can tell.

You can’t copy-paste experience. You can watch someone exercise great judgment for years and still not absorb it because the actual mechanism is invisible. Traditional organizations solved this through time. Apprenticeships. Mentorships. Knowledge transferred slowly through osmosis, through a thousand micro-corrections.

This is also why so many people struggle with async work. Synchronous collaboration lets you orient in real-time. Every raised eyebrow, every clarifying question, quick “no, not like that.” The tacit stays tacit because presence fills the gaps. Async removes the safety net. All the context, standards, and judgment have to be front-loaded into artifacts that travel without you. The same skills that let you orient AI effectively are the skills that make async work possible: the ability to articulate what you know before you’re asked.

But when AI can produce in minutes what used to take days the limiting factor is how fast you can orient it. And orientation requires making the tacit explicit.

How to orient

When you have tacit knowledge

You have context, standards, and judgment that you haven’t articulated. The work is surfacing it.

AI is useful here—not as an executor, but as an elicitor. It can interview you. It can propose framings and let you react. It can offer bad answers that provoke you into articulating the good one.

Before you ask for output, ask for questions. “I need to write a memo recommending we cut Project Atlas. Before you draft it, interview me. Ask about the history, the stakeholders, what I’m worried about, what success looks like.”

The questions don’t need to be sophisticated. The value is in forcing articulation before delegation.

When you don’t have tacit knowledge

What if you’re working outside your domain? You can borrow judgment.

AI has seen things. Thousands of strategic memos. Millions of code reviews. Countless project post-mortems. It has absorbed patterns even if it hasn’t felt the pain behind them.

You can access those patterns—not by asking for output, but by asking for orientation:

“What usually goes wrong with this kind of project?” “What would a senior engineer push back on here?” “What questions would an experienced investor ask about this pitch?” “What am I probably not thinking about?”

You’re not asking it to do the task. You’re asking it to surface the patterns from all the examples it’s seen, distilled into concerns and questions and failure modes. Borrow the orientation, then direct the output.

The crystallization loop

The failed outputs are more valuable than the successful ones.

The tacit only becomes visible when something misses in a way you feel but couldn’t have predicted. That feeling of wrongness—“this isn’t quite it”—is tacit knowledge surfacing. Most people treat this as friction. A cost. They refine the prompt, get a better output, and move on.

But there’s something more valuable available: you can use each failure to excavate the knowledge that made you dissatisfied in the first place.

When an output feels wrong, pause. Ask yourself: Why is this wrong? What standard did it violate that I never stated? What context was I holding that I didn’t share?

Each complaint is a principle waiting to be made explicit. Write it down.

“I don’t like this because it’s too formal for our brand voice.”

“This misses the point because cost isn’t the real concern—timeline is.”

“This would never work because it assumes we have exec buy-in, which we don’t.”

Crystallize it into an artifact to convert tacit knowledge into explicit orientation that can be reused.

The loop looks like this:

Orientation → Output → Dissatisfaction → Crystallize Artifact → Better Orientation

Every cycle through the loop adds to the artifact library. Your orientation gets richer. Your outputs get closer on the first try. The work isn’t getting a good output once. The work is building the orientation that makes every future output better.

Create artifacts

Most feedback is tactical and disposable. Specific to a single task. Orientation is structural and cumulates through artifacts to shape entire categories of work.

Artifacts are orientation made durable. They’re the container that lets your judgment travel. To future you, to collaborators, to systems that can apply them without asking.

They can take many forms.

Principles are the simplest. One-liners that encode a standard.

• Our voice is warm but not saccharine. No exclamation points in subject lines.
• Assume the reader is skeptical and busy. Lead with benefit, not product.
• We don’t compete on price. Never position against cheaper alternatives.

Rubrics add structure. They turn implicit evaluation into explicit criteria:

• A good proposal has: clear problem statement, quantified impact, named owner, timeline with milestones, explicit risks.

Briefs encode context for a specific domain:

• Project Atlas context: killed Q2 for timeline reasons, not cost. Key stakeholder (Maya) was the original champion—handle carefully. Exec team is skeptical of pivots right now.

Examples show rather than tell. A before/after pair. An annotated draft. A “this, not that” comparison.

The form matters less than the durability. Whatever you create should be something you can hand over without explanation—something that orients on your behalf when you’re not in the room.

Most feedback disappears. “Too formal.” “Wrong angle.” “Not quite.” The output improves; the feedback evaporates. The next task starts from zero.

Artifacts are what you get when you refuse to let that happen. When you treat every correction as a crystallization opportunity. When you pause after “not quite” and ask what principle your dissatisfaction reveals.

The correction fixes one output. The artifact fixes every output like it.

Accumulation

Run the loop long enough and something shifts.

At first, each task requires orientation from scratch. You’re doing the work of articulation repeatedly—explaining context, restating standards, surfacing judgment. It’s slow. It feels like overhead.

But the crystallizations accumulate. And, at some point, you cross a threshold: new tasks in familiar domains require minimal new orientation. You hand over the library of artifacts and the output lands close on the first try.

Then something stranger happens. The orientation becomes transferable. It’s no longer just you orienting—your artifacts begin to orient other people’s work, other systems, other minds entirely. The context that once lived only in your head now exists as an artifact that scales beyond you.

This is what it means to manage infinite minds.

Not steering each one. Not writing each prompt. But having built the orientation layer that steers them—a library of explicit judgment that can be applied by anyone, to any mind, at any scale.

The new work

Work is shifting. Seniority used to mean “produces better.” Increasingly, it means “orients better.” The senior engineer’s value isn’t in the code they write—it’s in the judgment they can crystallize and transfer. The experienced strategist’s value isn’t in the memo—it’s in the artifacts that make the memo land without revision.

Every time you feel dissatisfied with an output and take the time to ask why, you’re doing this work. Every artifact you create is a small act of scaling yourself.

Run the loop enough times and you stop managing outputs.

You start managing minds.