Choose the useful outcome
Keep the required outcome and acceptance conditions stable while examining which inherited means can change.
TRANSFORMATIONAL INTELLIGENCE ∞
Intelligence that accelerates
the means of acceleration.
Build the capabilities that make the next advance possible. Measure the time to accepted outcomes, including what it takes to get there.
The central idea
Some progress comes from doing a task faster. A deeper opportunity is to create an instrument, representation, production process, or voluntary arrangement that changes what the next task requires.
The next advance should help build the one after it.
OUT-ACCELERATE develops that idea as a research framework: diagnose the binding constraint, construct the missing means, qualify the result, and test what becomes possible on fresh successor work. Time saved must survive construction, verification, failures, maintenance, and financing.
When to reach for this paper. Use it when additional effort keeps meeting the same constraint, or when a reusable capability could shorten the route to a consequential accepted outcome.
Keep the required outcome and acceptance conditions stable while examining which inherited means can change.
Compare direct execution, reuse, and complementary changes. Count the time and resources needed to construct and qualify them.
Use strong complete alternatives, failures, and counterexamples to distinguish a useful capability from a faster isolated step.
Carry useful means into fresh work. Check their current scope, available resources, and demonstrated contribution.
This research edition presents a framework and an exact analytical worked example. The example illustrates conditional mechanisms under stated assumptions; it does not establish autonomous invention, measured frontier-model acceleration, physical deployment, or a commercial result.
Use the companion to inspect the model, reproduce its calculations, and check the comparison. The lab’s US$3 billion trailing-365-day complete-cost profit objective remains unachieved.
Read the flagship foundation · Explore the practical mechanism guide
Use the PDF controls to search, zoom, and change pages. On a phone or if the preview is blank, open the full paper or download the PDF.
The editable LaTeX, bibliography, and companion are available together in the public repository.
Boucher, Vincent. OUT-ACCELERATE: Intelligence that accelerates the means of acceleration. MONTREAL.AI, research edition v1.0, 2 November 2026.
Research edition v1.0 · 2 November 2026. The reading page is the stable entry point; a Git commit permalink identifies an immutable edition.
21 pages · 175,793 bytes. SHA-256: fbc59b486d7494ed97077adeaa81b6ee96bab5f27b3274ffd88df6edbf358669
The publication manifest records the exact edition alongside the other published papers.