EveMissLab
One lab. One site per thing it builds.
EveMissLab is a research and engineering lab. Everything it finishes gets published as its own site, under its own domain, with its own source repository — so each piece can be read, checked and used without going through the others.
The index
Eight sites live today. The list is built to keep growing.
Auditable computation
Built on one idea: a result is worth less than a result you can re-run, and a claim is worth less than a claim that states its own boundary.
- MMR-Benchmmr.evemisslab.comThree independent readings of every formula cell, signed Ed25519 computation certificates, and exact replay. Engine disagreement never authorizes a formula rewrite.
- MMLC Runtimemmlc.evemisslab.comAn auditable runtime for typed matrix-ledger documents: deterministic and symbolic execution, provenance, constraints, temporal fixed points, counterfactual branches and finite decision analysis in one execution model.
- MLFmlf.evemisslab.comAn AI-native matrix knowledge format with a reference compiler. Keeps coordinates, regions, roles, formulas, dependency edges and provenance instead of flattening them into one token sequence.
- DIEECdieec.evemisslab.comA symbolic runtime for dual internal–external expansion computation: a frozen public API, one certificate envelope for every claim, migration for old ledgers, and a release gate that treats passing tests as necessary but not sufficient.
- ANLAanla.evemisslab.comA lossless archive format an AI can plan and a deterministic, model-independent decoder must restore exactly. Two reference implementations, cross-verified byte for byte.
- MMPFmmpf.evemisslab.comA route-aware factorization runtime that has to choose before it knows what the choice will cost. Performance mode takes the lowest predicted cost; assurance mode takes the lowest bound it can certify, and is slower on purpose.
- MMRFmmrf.evemisslab.comA public prime dataset defined as much by what it refuses as by what it answers. The query surface is aggregate-only, and the guard that refuses a target-conditioned request runs before any shard is read.
- UTF-8Xutf-8x.evemisslab.comUTF-8 stays the verified semantic anchor while storage, search, editing and inference each get a reversible representation suited to their own costs. AI generates the strategy and is never required to decode.
Research archives
Corpora meant to be read by machines as readily as by people.
Working systems
Things that run, rather than things that are described.
- FELRAfelra.evemisslab.comA Python-first academic verification and visualization workbench for GCPR–RWL–FELRA, turning theory and data into reproducible computational evidence.
- Storyforgestoryforge.evemisslab.comAn English-first bilingual writing and reading platform for AI-authored fables, fairy tales and classic reinterpretations.
- AI Boardai-board.evemisslab.comA public machine-readable notice board for AI agents, search systems and cognitive-architecture research. It answers browsers in plain text too, because it was not built to be looked at.
- DRVSdrvs.evemisslab.comA corpus-agnostic, mostly client-side search engine. Instead of replacing your page with a result list, it dims what does not match and tells you why what is left is there.
- MS3Ems3e.evemisslab.comA multilayer nested spectral skeleton state engine. Compiles linear media into an identity skeleton, deformation layers and a spectral decomposition, then reconstructs from a state vector instead of a timestamp. The engine itself runs on the page.
How this works
One site per thing
A project that is finished gets a subdomain, a source repository and a page that explains it. Nothing important lives only as a paragraph on this page.
Open by default
The code is on GitHub under Apache-2.0, with tagged releases and verifiable checksums where the project produces them.
Boundaries travel with claims
Every project states what it does not establish, next to what it does. That list is part of the specification, not a disclaimer appended to it.
Built with AI, said plainly
These projects are built in collaboration with AI systems. Where that matters to how a result should be read, the project says so.
Contact
Everything published by this lab is reachable from the index above. Source repositories are linked from each project's own site.