FUTURE
MAP ROOM Research status

FUTURE MAP ROOM GLP RESEARCH / 2026

The mathematics of
what can still happen.

An independent computational research space studying viable reachability, partial observability, decision geometry, and the difference between what is possible, visible, knowable, and worth doing.

Explore the findings ↘Open the framework
08completed
formalization stages
64worlds in exact
active-sensing model
E3highest current
internal evidence grade
LIVE MODEL 00

Viable route field

COMPUTING
NOWABCDEF2 / 3 ROUTES VIABLE AT τ 48
FUTURE MAP ROOM · LOS ANGELES

Exact finite computation + synthetic simulation

External technical review pending

Not a physical theory

01 / RESEARCH GRAMMAR

One world.
Four declared layers.

GLP keeps the external system, the observer, the task, and the evaluation rule separate. Most category errors begin when one of these disappears.

  1. 01EnvironmentStates · transitions · constraints · disturbances
  2. 02ObserverChannels · beliefs · models · attention limits
  3. 03TaskGoals · actions · horizon · stakeholders
  4. 04EvaluationLoss · regret · viability · robustness
RESULT 06E3 · EXACT CONSTRUCTED

More information is not always
more decision value.

In an exact 64-world navigation model, the policy that reduced more total uncertainty produced 120.5% more route regret than the task-aware policy.

Enter the simulation →

02 / INTERACTIVE RESULTS

Claims you can
touch.

Each module changes one declared variable while preserving the rest. Toggle the models. Watch the category errors appear.

EXPERIMENT 06 / DECOY CONDITION / BUDGET 2

Salience is not decision value.

G1G2G3G4G5G6STARTROUTES
SELECTED OBSERVATIONBlack light → fluorescent decoy G6

Visibility-first

Posterior entropy4.286 bits
Route regret3.738
Optimal-route probability50.5%
This policy reduces more total uncertainty, yet incurs 120.5% more regret.

Exact enumeration of 64 latent worlds. Standard Bayesian decision theory. The benchmark is reproducible; it is not a new POMDP theorem.

03 / TYPED CONTRIBUTION

One score is
not enough.

The first computational stress test rejected a master “possibility expansion” number. Interventions act on different mathematical properties, so the surviving proposal is a typed profile.

VRC = (ΔO, ΔN, ΔR, G⁺, D⁻, C, U)
Δ outcome0.000
Δ navigation-0.167
Δ robustness-1.000
Gross gain+3.000
Destructive loss+3.000

Equal gain and destruction cancel to zero. The structure changes radically.

Finite graph · horizon H = 5 · declared outcome map · explicit baseline

04 / EVIDENCE REGISTER

Show the work.
Label the limits.

Evidence grades describe internal computational strength, not empirical generality. Exact solutions to designed examples are still designed examples.

IDStudyMethodGradeSurviving result
GLP-001Viable Reachability ContributionExact finite graphE3Master scalar rejected
GLP-002Balance and Disagreement50,000-trial simulationE2Consensus separated from accuracy
GLP-003Distance, Access, RemotenessExact 504-target gridE3Model distance separated from physical distance
GLP-004Finite AttentionSynthetic sensing budgetsE2Value indexed to task and loss
GLP-005Selective IlluminationSynthetic channel modelE2No master visibility scalar
GLP-006Active IlluminationExact 64-world enumerationE3Salience separated from decision value
GLP-IG1Task-Relative Belief GeometryExact channel comparisonE3Equal information, unequal decision effect
GLP-IG2Compressive ConsensusExact six-agent simulationE3Provenance required for valid fusion

Technical notes, source code, figures, and hashes exist in the canonical record. Public downloads remain held until authorship, IP, and external-review gates are cleared.

05 / PROPOSED EXPERIMENTS

Designed to
fail cleanly.

The next phase moves from exact constructed models toward preregistered physical and human-observation tests. Each proposal includes the result that would weaken it.

PREREGISTRATION DRAFTP-01

Does the salience-consequence separation survive when people inspect real material markers under selectable illumination before choosing a route?

MINIMUM DESIGN
  1. 1Controlled route board with uncertain gates
  2. 2Black light, red light, and diffuse illumination
  3. 3Marker placement randomized between high- and low-consequence gates
  4. 4Measure route regret, inspection order, response time, and confidence
RELEASE GATE

Human-participant review, calibrated sensors, simulator-sickness not applicable

DISCONFIRMATION CONDITION

Visibility-first and task-aware policies perform equivalently across the decoy intervention.

06 / APPLICATION LAYER

PRIVATE BETA · MODULE RESERVED

Research that
can wiggle.

Wiggle turns the most defensible GLP distinctions into a transparent decision map. It does not predict fate. It shows what can move now, what needs prerequisites, what is blocked, and what still needs to be learned.

Possible nowAfter prerequisitesBlocked for nowUncertain
The current working beta will be integrated here after its latest build is reviewed. No personal decision data is used on this research site.
WIGGLE / DECISION MAPLOCAL MODEL
GOALChoose a filming location3 paths still have wiggle
GOALRANCHSTAGEHOUSE
WHAT TO LEARN NEXTConfirm after-hours accessCould reclassify 2 options · Low effort

07 / SUPPORT THE RESEARCH

Help keep the
room open.

Future Map Room is seeking serious partners before opening a public donation channel. Support begins with scrutiny, infrastructure, and aligned work—not a wallet address.

DOMAIN PILOTSS-01

Institutional partners

Universities, laboratories, public-interest groups, and production systems with a bounded decision problem suitable for a transparent pilot.

01External review02Bounded domain pilots03Shared grant work
PARTNERSHIP INTAKEContact channel openStart a conversation →
PUBLIC DONATIONSPaused
TAX-DEDUCTIBLE STATUSNot claimed
CRYPTO ADDRESSESNone published
NEXT RELEASE GATEEntity + fiscal sponsor

No contribution is currently solicited or represented as tax-deductible. Funding infrastructure will be opened only after custody, accounting, and review procedures are established.

08 / RESEARCH STATUS

Ambitious.
Falsifiable.
Unfinished.

Credibility is not a tone of voice. It is the visible separation between exact result, conjecture, interpretation, and metaphor.

SUPPORTED INSIDE DECLARED MODELS
  • Viable paths differ from merely feasible endpoints.
  • Consensus can be accurate, biased, or brittle.
  • Physical, operational, and estimated distances can invert.
  • Equal information quantity can have unequal decision value.
  • Provenance-free consensus can converge to a wrong policy.
NOT ESTABLISHED
  • A new branch of mathematics.
  • A universal possibility score.
  • A geometry of physical spacetime.
  • Attention changing external physical law.
  • Empirical generality beyond the tested models.
CURRENT PHASEVersion 2.1 internal red-team complete

Independent technical review remains outstanding. The program is in a bounded seed phase.

PRIOR ART POSITIONIntegration and reporting proposal

Core machinery comes from viability theory, control, graph algorithms, Bayesian decision theory, POMDPs, information geometry, and robust statistics.

AI ASSISTANCEDisclosed and reviewable

OpenAI systems assisted with drafting, formalization, code, analysis, and visual production. AI fluency is not treated as mathematical evidence.

DIRECTIONCasey Alexander

Founder and research director · Independent program · Los Angeles, California