Projects · Current focus · Ecosystem · Research output · Standards · Collaboration

Independent researcher and systems engineer at the Anulum Institute in Switzerland. The programme runs on one engine: an autonomous coding and verification system that builds, tests and reviews the projects on this page and binds their claims to evidence. The engine is put to work in four domains: reactor systems, systems integration and control, neuromorphic computing, and quantum computing. Claims are only as good as the measurements, artefacts, or verification that support them.

Projects for the software and its evidence · Papers for every publication and software archive with BibTeX · Contact for a technical proposal. Cold readers: anulum.li/start/ picks an entry point by profile. Every registered project, filterable by group, reactor family and evidence, with a clickable map: anulum.li/portfolio/.

Each card links to inspectable artefacts, not summary claims. Evidence links are pinned to the commit they were verified at. Release and CI badges report registry and workflow state; they are operational signals, not scientific-quality scores.

Synapse Channel · Usable now Control plane for coding-agent fleets: claims, roles, durable mailboxes, receipts, audit, and federation. Documentation · Validation · Coordination specification · Threat model	Director-AI · Research active Real-time LLM guardrail: NLI/RAG grounding, claim review, native acceleration, optional claim-level streaming halt, declared capability boundaries. Documentation · Validation · Public benchmarks · Capability matrix
Rigor Foundry · Usable now Evidence-bound repository inventory, audit candidates, review binding, and remediation planning. Documentation	Remanentia · Usable now Auditable memory for AI agents and knowledge systems: hybrid retrieval, graphs, consolidation, CLI, MCP, and API surfaces. Documentation
SC-NeuroCore · Research active Stochastic and neuromorphic framework: Python models, Rust SIMD paths, Verilog RTL, HDC/VSA, compiler surfaces, and hardware evidence. Documentation · Validation · Synthesis results · Traceability matrix	SCPN Fusion Core · SCPN Control · Research active Tokamak physics, solvers, and validation campaigns with real-data paths (Fusion Core); control-grade runtime with fail-closed admission and replay evidence (Control). Validation · DIII-D validation record · Software DOI
SCPN Quantum Control · Experimental Evidence-governed quantum simulation of coupled-oscillator synchronisation: protocols committed before the run, hardware result packs, raw counts, and explicit non-advantage boundaries. Layout-relaxation protocol · Result-pack contract · Software DOI	SCPN Phase Orchestrator · Research active Evidence-first synchronisation analysis and review-only control proposals for coupled rhythmic systems; matched-false-alarm evaluation, negative results, and bounded transfer claims. Submission index · Negative-result preprint · Grid regime-map preprint

_{Portfolio state verified 2026-09-25.}

_{Rendered from anulum.li/news (the projects' CHANGELOG files) and Zenodo on 2026-09-25.}

_{Rendered from anulum.li/news (the projects' CHANGELOG files) and Zenodo on 2026-09-25.}

Timeline entries distinguish registry facts from self-published chronology. They do not imply academic affiliation, external validation, funding, or awards.

flowchart LR
    A["01 · Inspect<br/>Rigor Foundry"] --> B["02 · Guard<br/>Director-AI"] --> C["03 · Coordinate<br/>Synapse Channel"] --> D["04 · Compute<br/>SC-NeuroCore · SCPN"]

Research, validation, and product readiness remain separate. Active development is not a readiness claim.

The map contains 39 portfolio repositories: 33 public repositories and six private product surfaces. HushLine is a standalone public utility outside the five portfolios. Connections represent contract, integration, evidence, and audit flow. They do not merge ownership or imply scientific validation, operational readiness, or actuation authority. The interactive version of this map, with every registered project as a row that filters by group, reactor family, visibility, lifecycle and evidence, is anulum.li/portfolio/; each portfolio below links to its own view of it. Counts verified 2026-09-25. The account lists more public repositories than the map: the 34 mapped public projects plus this profile repository and a few forks kept for reference.

Access key: PUBLIC · PUBLIC / ARCHITECTURE-ONLY · PRIVATE · PRIVATE / PROPRIETARY · PLANNED (plan written, no repository yet)

Open work here is funded through GitHub Sponsors. About USD 25,000 a month is the minimum that keeps the lab running without interruption: compute, GPU, QPU and FPGA time, hardware, CI and tooling. This is lab cost, not a salary. Above that minimum the work scales, because the engine turns additional compute directly into more built and verified work. Sponsorship pays for the lab; it does not buy results, and findings are published whichever way they fall.

To direct your support, name the engine or one domain when you sponsor; the money is spent there and reported in the quarterly note. The hardware the lab needs is listed item by item in LAB_EQUIPMENT.md, and equipment can also be given in kind.

03 · Agentic Coordination, Assurance and Continuity. The autonomous coding and verification system that builds and checks every project on this page, taken to 1.0 as open infrastructure. It combines SYNAPSE CHANNEL (agent coordination with receipts), Rigor Foundry (evidence-bound review), Director-AI (grounding of claims) and Remanentia (auditable memory). Sponsorship pays for model and GPU time for the agent fleet, evaluation benchmarks and CI.

• Research agreement. An organisation can fund a defined piece of work under a written research agreement, invoiced directly, instead of a sponsorship. Scope and deliverables are agreed first; findings are published under the same rule as above.
• Commercial licence. Projects under AGPL-3.0-or-later can be licensed on commercial terms for proprietary or closed use, for example SYNAPSE CHANNEL, SC-NeuroCore, SCPN Control, SCPN Fusion Core, SCPN Phase Orchestrator, SCPN Quantum Control. Apache-2.0 projects such as Director-AI and Rigor Foundry need no licence; Director-AI Pro is a commercial self-hosted edition, and support and integration can be contracted for any project. The LICENSE file of each repository states which licence applies.
• Contact: protoscience@anulum.li, with "research agreement" or "commercial licence" and the project name in the subject.

The verified PyPI profile currently contains 19 published projects. They include public Python packages, Rust-accelerated engines, domain kernels, and command-line tools.

Primary implementation

Practices are applied according to each repository's risk and scope. Not every repository runs every tool.

Negative and null results are published when they are real. Public claims stay tied to measurements, protocols committed before the run, raw packs, source-bound reviews, or executable verification.

Example: SCPN Quantum Control publishes protocols committed before the run and hash-bound result packs rather than converting experimental activity into a readiness or advantage claim.

• Evidence before claims.
• Reproducible artefacts before presentation.
• Clear boundaries between research, validation, and product readiness.
• Cross-language implementations where performance or hardware integration justifies them.
• Fail closed when provenance, authority, or evidence is incomplete.
• Negative results and failure records remain part of the research output.

I welcome technically grounded collaboration in reliable AI infrastructure, multi-agent systems, neuromorphic computing, scientific software, formal verification, quantum simulation, plasma physics, and control.

Engagements are taken selectively. The public GitHub profile currently marks the account as hireable; this is not a guarantee of immediate capacity.

A useful first message includes the problem, constraints, relevant prior art, and what evidence would count as success. Contact me through protoscience@anulum.li or anulum.li.

I respond to technical proposals. I do not take on ungrounded hype work, demo-only science theatre, or claims that cannot be checked.

For sustained open work, GitHub Sponsors funds CI runners, compute, hardware and quantum experiment time, and public documentation rather than marketing.

Transparency: These repositories span research software, developer tools, private systems, and product candidates. Active development does not imply production readiness or scientific validation unless a project provides explicit evidence.

I AM THAT