Mike Litman
Output Stacking
A TALK BY MIKE LITMAN

Output Stacking

One person, three sessions, and a stack of briefs replaced the old shape of work.

I used to do one thing at a time.

Then I opened a second window. Then a third. Then I noticed.

01
ACT ONE

Pre-AI, work was a queue

One task in front of you. Deep focus when you could get it. Multitasking was a known cost: every switch carried attention residue, every interruption rebuilt context from scratch.

You were the bottleneck because you were the only worker. The discipline was sequential. Single thread, single mind.

SERIAL WORK YOU = BOTTLENECK SINGLE THREAD

Then the bottleneck moved.

From hands to attention.

02
THE INFLECTION POINT

Stacking became possible the moment AI could run unsupervised

Pre-2024, AI was autocomplete. Each turn needed a prompt, a check, a correction. You couldn't leave one running while you briefed another, the babysitting tax ate the gain.

Then sessions got capable enough to take a real brief, ask the right questions, and ship work without constant supervision. The moment one session could hold its own for an hour, the second session became free. Then the third. Then the stack.

AGENT-CAPABLE SESSIONS RUNS UNSUPERVISED THE STACK BECAME FREE
03
THE SHIFT

You open a second session. Then a third.

One window for the deck. Another for the email replies. A third spinning on the data audit.

Each one running an isolated worker that doesn't get distracted, doesn't get tired, doesn't context-switch. You moved from worker to orchestrator without noticing it had happened.

3+ PARALLEL SESSIONS NO CONTEXT SWITCHING YOU = ORCHESTRATOR
04
THE TRAP

This looks like multitasking. It isn't.

Multitasking is what happens when one mind tries to hold two tasks at once. It fails because attention is finite: every switch leaves residue, every return rebuilds context. The research is decades deep.

People who multitask don't get more done. They do everything worse.

ATTENTION RESIDUE · LEROY 2009 SWITCH COST WORSE OUTPUT
THE DISTINCTION

Multitasking is bad because you fragment.
Stacking works because the AI doesn't.

Then
Multitasking: one mind, many tasks

The same person tries to hold conflicting context. Quality drops on every track. Fragmentation is the cost of trying.

Now
Stacking: many minds, one orchestrator

Three deep-focused workers run beneath a manager who switches between them. Each session is its own runtime, with its own state and no cognitive crosstalk. Fragmentation only at the briefing layer, never inside the work.

WHAT THIS LOOKS LIKE IN PRACTICE

My stack, right now

🪟
3 sessions live
Voice agent ops in one. Content pipeline in another. Product builds in a third.
📋
Scoped CLAUDE.md per project
Each session reads only its own rules. No leak across lanes.
📖
Project bibles as briefs
One bible per product. The brief replaces the conversation.
🧭
START HERE.md
Daily orientation. Each session reads it cold. Zero re-briefing.
Cron jobs run while I sleep
Newsletter, audits, monitors. Sessions that don't even need me awake.
🤖
Claude as orchestration layer
I write briefs. It runs the workers. I review the outputs.
05
THE SYNC MECHANIC

Sessions don't talk. You do.

Three sessions can't read each other, that's the safety property. But three sessions building parts of one deliverable need to know what the others have done. Without that, the stack drifts.

The fix is the session update: one session writes a summary you can paste into the next. State moves across lanes through the orchestrator, never directly. Sessions stay isolated. Context still flows.

This is the routing layer. Without it, three sessions are three orphans. With it, three sessions are one coordinated deliverable.

SESSION UPDATE ROUTING THROUGH YOU ISOLATION + SYNC
06
THE DISCIPLINE

Stacking only works when you brief like a manager

Each session needs a rigid brief: what it owns, what it doesn't touch, what 'done' looks like. Like a one-person agency hiring three contractors at the same time.

The brief replaces oversight. The brief is the work now.

SCOPED BRIEFS OWNED LANES DONE-CONDITIONS
07
WHAT BREAKS

Lane leaks. The moment a session crosses lanes, output collapses.

Leaks happen when briefs are loose. Three sessions writing different parts of a single deliverable, all needing to converge cleanly at the end. If the brief doesn't draw lanes hard enough, the second session starts overwriting work the first nearly finished, not because it's wrong, because it doesn't know to stay out. Version control turns from a backup into a battleground.

The fix isn't smarter sessions. It's tighter lanes: which session owns which file, what's locked, what 'done' looks like, what touches the merge step. Without that structure, output is worse than serial.

LANE LEAK OVERWRITE RISK DISCIPLINE = STABILITY
08
THE QUALITY OBJECTION

Yes, but is the work any good?

The most-asked question. The honest answer: yes, often better. Output quality with a stack tracks the quality of your briefs and your review layer. Both improve under stacking, because the brief now does the work the conversation used to do.

A loose brief produces loose output, just like a loose colleague. A tight brief, applied to a focused session that does nothing else, beats a tired sole executor on most of what knowledge work actually is. Quality lives in the brief and the review, not in who types the words.

BRIEF = OUTPUT QUALITY TIGHT BRIEFS WIN REVIEW IS THE GATE
09
THE LOAD QUESTION

But isn't orchestrating exhausting?

The most-asked follow-up. Multitasking exhausted you because one mind held multiple tasks in suspension, paying attention residue on every switch.

Orchestration is different. You write a brief, hand it off, step away. You don't sustain three internal models, you sustain one decision at a time: which lane, which brief, what's missing in the output you just got back. The load is bandwidth allocation, not deep-context holding.

The cost is real but tradable. A tighter brief now saves an hour of rework later. Orchestration scales with practice the way reading speed does, faster the more you do it.

BANDWIDTH ALLOCATION NOT CONTEXT HOLDING SCALES WITH PRACTICE
10
THE NEW ROLE

You're not a worker anymore. You're a one-person agency.

The sole executor died without ceremony. The new role looks like running a tiny consulting firm: write briefs, review outputs, decide what ships.

This isn't management with new tools. The latency is seconds, not weeks. The cost per worker is near zero. The orchestrator is also the worker when she wants to be. The feedback loop is so tight that briefing becomes thinking. The labour moved. The decisions stayed.

SOLO AGENCY NOT MANAGEMENT BRIEFING = THINKING
11
THE NEW SKILL

Briefing replaces doing

The differentiated skill of the post-AI worker isn't typing faster, prompting better, or knowing more frameworks. It's writing a brief tight enough that three sessions deliver three coherent outputs without colliding.

Editorial discipline applied to your own labour.

EDITORIAL DISCIPLINE TIGHT BRIEFS NO COLLISIONS
WHAT THIS MEANS

Most people are still queuing tasks. Some are stacking them.

For knowledge workers
Your output ceiling expands the moment you can stack. The bottleneck is your attention, not your hands. The compounding gap between stackers and queuers is already opening.
THE BET

By 2030, output stacking will be the default working pattern for most roles that produce knowledge.

Some work stays sequential by nature: the live call, the negotiation, the in-the-moment decision. Stacking is for everything else, which is most of it. Sequential focus will become the boutique mode. The skill will be in how you brief it.

A TALK BY MIKE LITMAN

Multitasking failed because you fragment.
Stacking works because the AI doesn't.

mikelitman.me · hello@mikelitman.me

Continue from slide ?