Constraint Dynamics  /  Funding
Open-source launch · geometric knowledge · community review

Fund geometric knowledge
for safer AI discovery.

Golem Physics is already a working research system with preserved runtime evidence. Funding now supports the responsible open-source launch: repository cleanup, reproducible setup, curated evidence release, benchmark harness, community review, and the Constraint Native proof walkthrough that carries verification before voice into governed action.

Request the funding packet → Evidence map Open Golem
01 · The case

Golem leads the research.
Constraint Native carries the bridge.

The work is a safety research program, not only a better MCP gateway. The thesis is concrete: claims become coordinates, geometry reveals what is missing or conflicted, dream cycles test what should crystallize, and speech waits for status. The next phase is release and review, not a request to invent the system.

01 · Grant

Golem is the research system.

The funding round turns the working Golem Physics artifact into a public release path: installable code, curated evidence, benchmark harness, dataset, and evaluation notes for claim status, abstention, proposal leakage, provenance retention, and false crystallization behavior.

02 · Evidence

The system runs today.

Dated lattice metrics, preserved runtime evidence, an inspectable app with discovery geometry, an adaptive dream-cycle control loop, the DOI-backed Golem release, and published conceptual lineage behind the work.

03 · Bridge

Constraint Native translates the discipline.

The agent firewall is the buyer-facing surface: governed action, policy gates, approvals, taint tracking, signed proof paths, and offline replay.

02 · The ask

Open-source launch first.
Bridge-ready underneath.

The preserved Golem Physics bundle already supports diligence. The near-term work is to make it usable by others: repository cleanup, public installation notes, a reproducible demo bundle, curated evidence release, benchmark harness, and reviewer walkthroughs that connect Golem to Constraint Native without broadening the claim.

Track A · Open-source research launch

Research audience, release framing.

Track A is the open-source path: Golem as the implemented testbed for verification before voice, abstention, provenance retention, proposal separation, contradiction preservation, and false crystallization measurement.

The ask is time and focus: clean the repository without destroying provenance, document installation, prepare a reproducible demo bundle, curate public evidence, and specify the benchmark without broadening the claim.

Vehicle · grant / stipend / research support Audience · AI safety reviewers Output · release, dataset, evaluation harness
Track B · Bridge funding

Organization runway, $50-100K bridge.

Track B is the practical bridge: a grant, microgrant, or strategic check that keeps the research organization moving while the Golem release is prepared and Constraint Native proof artifacts stay current.

The packet is sized for diligence: use of funds, 12-month budget, Golem evidence map, proof sample, release plan, and direct founder contact at matthewacator@gmail.com.

Vehicle · grant / microgrant / strategic check Amount · $50-100K Output · grant submission plus funder-ready packet
03 · What exists now

Working proof,
not a deck.

Funding accelerates the public release of what is already visible: the Golem app, dated lattice metrics, preserved runtime evidence, geometric rediscovery results, dream-cycle behavior, the DOI-backed Golem release, the conceptual lineage behind it, and the macOS-first governed agent boundary. The next dollar makes the artifact easier to run, inspect, benchmark, and review, not the story broader.

evidence funding-readiness current
Golem appClaim Studio, Evidence Cockpit, Discovery Engine, Hypothesis Engine, Anomaly/Analogy, Expectations, Silence Map, Lattice Graph, Oracle Chat, Activity River, Dream Theatre.working
Geometric architectureClaims carry coordinates, embeddings, domain placement, provenance, status, tension state, volatility, immutability, and temporal history.lattice
Runtime lattice snapshot13,240 lattice nodes loaded · 215 immutable nodes · 91 tensions loaded · 21 proposed nodes · 4,035 experience nodes · 43 domains sampled in discovery scans.observed
Geometric rediscoveryProof-ready run: 20 discovery cards filtered to 9 evidence packets, with 5 known-structure hits across physics, logic, mathematics, and epistemology neighborhoods.milestone
Benchmark planRelease deliverable: benchmark harness, dataset, and evaluation notes for false crystallization, proposal leakage, abstention precision, provenance retention, and contradiction preservation.planned
Discovery engineActive evidence seeking, crystallization of supported claims, rejection of unsupported proposals, preserved tensions, and recycled unresolved contradictions across cycles.engine
Dated evidence caveatLive counts and app captures should be re-run and labeled clearly before public release, grant, fellowship, or reviewer submission.refresh
Adaptive control loopMutation timeline shows dream-threshold adjustments, embedder retrains, and coherence-floor moves with named causes.control
Primary paperThe Outlines of Sanity: Constraint-Native Inference in Minds and Machines · DOI 10.5281/zenodo.19658730 · published 2026-04-20.Golem
Conceptual lineageThe earlier Constraint Dynamics Zenodo record remains background theory; it is not the evidence lead for this funding path.lineage
Commercial bridgeConstraint Native — governed coding-agent boundary with signed proof paths.bridge
Proof shapeSample signed proof path: 11 events · 4 policy blocks · 2 quarantines · 1 denied write · 1 denied network · 1 canary block · ed25519 chain.sample
Reviewer artifactsSample proof JSON, blast-radius report, replay-audit walkthrough.artifacts
Supporting architecturePersonal Mirror / OMP · OCP · Open Context Protocol · Resonance Region · proposed Golem Network.stack
Near-term workClean public repository path, curate release evidence, build a reproducible demo bundle, rerun lattice metrics, and polish the follow-on packet.next
claim discipline: bounded · local-first · proof-led release-ready after curation
JSON · 5KB

Sample signed proof path

The shape funders will see in a fresh run: governed events, policy outcomes, ed25519 chain, and evidence boundaries.

Open sample → MD · blast-radius

Blast-radius report

Per-session bounded-claim report: files touched, network attempts, secrets exposure, denied actions.

Open report → MD · replay

Replay audit walkthrough

How a reviewer walks the chain offline, verifies signatures, and reproduces the policy decisions.

Open audit → MD · Golem

Golem evidence map

A concise summary of the grant metrics, proposal separation, benchmark plan, and evidence boundaries for outside reviewers.

Open map → MD · rediscovery

Geometric rediscovery milestone

The latest proof-ready run shows Golem's geometry recovering known scientific structure and producing auditable candidate packets.

Open milestone → Zenodo · Golem

Constraint-Native Inference in Minds and Machines

DOI 10.5281/zenodo.19658730 · published 2026-04-20. Two PDFs: Part I covers Golem; Part II develops verification-first AI under constraint.

Open record → HTML · evidence

Current Golem evidence page

Use this current page before the older PDF packet is refreshed; it carries the app, Zenodo, paper, metric, runtime-evidence, and benchmark caveat story.

Open evidence →
04 · Reviewer artifact

A clean reviewer bundle
already exists.

A clean reviewer-facing Golem Physics bundle already exists, separate from the live lab workspace. The bundle is a copy-only artifact: useful for diligence because it keeps source, notes, runtime evidence, validation summary, manifest, and checksums in one bounded review surface.

Source

Copied operational source.

The reviewer surface includes copied source and grant-facing docs without asking reviewers to navigate the live lab workspace.

Notes

Curated Obsidian export.

The long-form design record is carried into the packet as curated notes, so the theory, architecture, and reviewer trail stay readable.

Evidence

Runtime evidence preserved.

Runtime evidence, validation summary, manifest, and checksums make the package inspectable without turning the public website into the artifact itself.

reviewer-bundlecopy-only artifact available on request
BoundaryLive lab workspace kept separate from reviewer evidence; the bundle is a bounded copy for review, not the runtime itself.separate
ContentsCopied source, curated Obsidian export, grant summaries, full runtime evidence, validation summary, manifest, and checksums.packet
What it showsPackaging discipline and inspectability: the work is not only a pitch deck, and reviewers have a clear artifact path.discipline
What funding doesRedact sensitive lab paths, refresh metrics, curate the public slice, produce a reproducible demo, package the benchmark harness, and prepare the release for community review.release
preserved evidence exists · public release after curation request access via funding packet
05 · Use of funds

Turn the working system into
shared research infrastructure.

The immediate work is not to invent the research program or prove that a system exists. It is to open-source the working artifact responsibly, preserve provenance, make a small reproducible path for reviewers, and let outside researchers test whether the architecture deserves follow-on support.

fundingwhat funding unlocks $50-100K
Open-source launch$50-100K funds repository cleanup, install docs, reproducible demo bundle, curated evidence release, benchmark harness, fresh Constraint Native proof export, and community review.unlock
0-10 daysFreeze release scope, redact sensitive paths, update grant/fellowship packets, and publish a clear release roadmap.scope
10-30 daysFresh Golem metrics, curated runtime evidence, reproducible demo bundle, and Constraint Native proof export.release
30-60 daysBenchmark harness, reviewer walkthrough, public install path, and curated evidence slice.bundle
60-90 daysCommunity review, follow-on grants, external technical review scope, and commercial bridge polish.bridge
Track 01

Open-source launch package

Prepare the public release without flattening the research record: license choice, dependency review, repository cleanup, install notes, first-run validation, and a small demo bundle reviewers can run.

→ release path
Track 02

Curated Golem evidence release

Prepare reviewer-safe walkthrough notes and screenshots for Claim Studio, Evidence Cockpit, Discovery Engine, Hypothesis Engine, Anomaly/Analogy, Expectations, Silence Map, Lattice Graph, Oracle Chat, Activity River, Dream Theatre, mycelium/temporal memory, and the mutation timeline.

→ inspectable
Track 03

Benchmark plan

Build the benchmark harness: specify false crystallization, proposal leakage, abstention precision, provenance retention, temporal correctness, contradiction preservation, and useful-answer rate as measurable targets. Produce the dataset, evaluation notes, and first baseline comparisons.

→ evaluation plan
Track 04

Theory and paper appendix

Present the published Constraint Dynamics theory and the Golem release as the conceptual foundation. Keep implemented single-system claims separate from proposed network extensions.

→ credibility layer
Track 05

Foresight and NSF follow-ons

Adapt the packet for Foresight AI for Science & Safety and NSF SBIR. Shift emphasis from existing artifact to reproducible infrastructure, open-source review, and commercialization.

→ non-dilutive capital
Track 06

Constraint Native commercial bridge

Keep the agent firewall materials current as the commercial bridge: demo, proof export, threat model, and buyer-readable positioning that follows the research proof.

→ commercial path
06 · Funder order

Submit in the order
the case is sharpest.

The first application is narrow and research-led. Later paths widen toward AI-for-science, safety infrastructure, cybersecurity, and commercial translation.

Target · 01

AI safety grant or fellowship

Golem as safety evaluation: abstention, provenance, proposal leakage, contradiction preservation, false crystallization, and benchmarkable behavior — implemented today, made reproducible through the grant work.

Target · 02

Foresight AI for Science & Safety

Golem as epistemic infrastructure for AI-assisted science: hypothesis generation, anomaly surfacing, and analogy bridging without promoting proposals into verified belief.

Target · 03

NSF SBIR / STTR

Translation into trustworthy AI infrastructure with commercial potential: reproducible evaluation, customer discovery, and product hardening for Constraint Native.

Target · 04

Cybersecurity grants

Constraint Native as agent-era endpoint governance: prompt-injection containment, local policy gates, signed proof paths, and offline replay.

Target · 05

Innovation funders and microgrants

The same evidence packet underwrites runway: app proof, grant calendar, use-of-funds memo, demo artifacts, and commercial-bridge plan.

Target · 06

Strategic infrastructure backers

The longer conversation: Golem as verified memory, Constraint Native as governed action, and the wider verification-first architecture.

07 · 90-day outcomes

A release plan with
concrete artifacts.

The success condition is concrete: a reviewer can understand the implemented system, run a bounded demo, inspect dated evidence, evaluate the benchmark harness, and see how Constraint Native carries the same discipline into governed action.

plan90-day · 2026-Q2/Q3 tracking
0-10 daysFreeze release scope, update funding packets, redact sensitive local paths, refresh live metrics, and publish the open-source launch roadmap.→ release scope
10-30 daysPrepare the public repository path, reproducible demo bundle, curated runtime evidence, evidence map, and fresh Constraint Native proof export.→ release bundle
30-60 daysBuild the benchmark harness and reviewer walkthrough while adapting the packet for Foresight, NSF SBIR, and other non-dilutive follow-ons.→ benchmark
60-90 daysInvite community review, scope external technical review, submit follow-on grants, and keep the commercial bridge materials current.→ review
scope: artifact-led, not promise-led commit: 2026-04-25
08 · Claim discipline

What this round is, and is not.

Stated explicitly so reviewers do not have to read between the lines.

In this round

  • Preparing the responsible open-source launch around Golem Physics
  • Cleaning the public repository path without destroying provenance
  • Refreshing live lattice metrics and curating release-safe runtime evidence
  • Building the benchmark harness and reviewer walkthrough
  • Refreshing the Constraint Native signed proof path and replay artifact
  • Scoping external technical and security review
  • Updating one-pager, memo, budget, and follow-on funding packet

Not in this round

  • Compliance certification (SOC 2, FedRAMP, HIPAA, etc.)
  • Replacement for IAM, EDR, DLP, secure-code-review, or human judgment
  • Perfect prompt-injection containment guarantees
  • Full endpoint isolation or hypervisor-grade sandbox claims
  • Cross-platform GA — alpha-scope until the proof is reproducible
  • Clinical, medical, or consciousness claims; Golem is not a chatbot product
  • A full institutional venture raise as the first public ask
Funding packet · request

Review the Golem evidence,
the commercial bridge,
and the use of funds.

The packet, sized to the funder: the Golem evidence map, one-pager, use-of-funds memo, 12-month budget, funding strategy, support-readiness checklist, and a bounded Constraint Native proof sample.

Request the funding packet → Read the architecture