MONTREAL.AI · QUEBEC.AI
Forecasting a World That Accelerates
An Evidence-Grounded, Constraint-Limited, Regime-Switching Framework for AI-Era Decisions
Realized outcome = residual bottleneck × min(technical capacity, absorptive demand)
Compact canonical model
Forecast realized outcomes—not capability in isolation.
Technical possibility becomes consequential only when it survives reliable execution, verification, authority, adoption, and the bottlenecks of the physical and institutional world.
Ys(t) = Bs(t) min { Q0 exp[∫0t gQ,s(u) du], Ms(t)Ds(t)Us(t) }
Realized output is the lesser of what technology can reliably produce and what the relevant market, institution, or mission can absorb, after remaining real-world bottlenecks are applied.
Version 2.0.0 retains the transparent hard-min equation as the primary model and adds a testable generalized family with separate technical, demand, and transfer bottlenecks plus optional partial substitution.
The formal paper
Read the complete contribution in the browser.
The site publishes the exact version-controlled PDF automatically from the repository whenever the paper or publication site changes.
- Canonical and generalized realization models
- Growth, acceleration, saturation and regime transitions
- Task graphs, reliability, verification and authority latency
- Probabilistic model averaging, hindcasts and ablations
- The complete Acceleration-Aware Forecasting Protocol Ω
- Reproducibility, governance, licensing and falsifiability
Paper SHA-256: 9eb050bb71b867ad08d3fb5b7b820f73433ce1a5ff83bc3327a18a30ee974c9f
The missing translation
Six layers that should never be collapsed.
The framework explicitly separates what systems can do from what institutions can authorize, absorb and realize.
Capability
Reliable execution
Verified completion
Authorized deployment
Adoption
Realized outcome
Coordinated research system
Formal theory, executable protocol, machine contract and validation program.
01
Formal paper
Equations, derivations, candidate regimes, uncertainty, decision rules, limitations and falsifiability.
Read the paper →02
Complete protocol
The Acceleration-Aware Forecasting Protocol Ω appears in the paper and as a standalone operational artifact.
Open the protocol →03
Machine contract
A versioned specification and strict schemas make the protocol auditable and interoperable.
Inspect the specification →04
Reference forecast
A complete generated report demonstrates scenarios, intervals, milestones, triggers and actions.
Open the report →05
Historical demonstrations
Source-linked METR and SWE-bench cases exercise unbounded, bounded, change-point and saturation behavior.
Open METR → · SWE-bench →06
Release integrity
The live site includes artifact hashes, build provenance and the exact source commit.
Checksums → · Build record →Verified release profile
Measured quality gates, explicit evidence boundaries.
The bundled release distinguishes mechanical verification and bounded demonstrations from prospective validation that must still be earned.
56automated tests in the frozen local release
90.54%branch-aware coverage across the declared scientific core
53paper pages, including the complete final prompt
4 regimesdownside, base, accelerated and discontinuous
What is supplied
Formal semantics, source code, schemas, controlled tests, historical demonstrations, deterministic artifacts, checksums and a prospective registry architecture.
What is not pre-claimed
Perfect accuracy, universal validity, peer review, independent replication, calibrated discontinuity probabilities, legal approval or guaranteed outcomes.