AxonOS

axonos.org · DY-WCET · Radar · DY Research · DY PROOF · Specifications · Articles

AxonOS is the hard real-time layer between neural hardware and the applications that use it: an open-source kernel in #![no_std] Rust on ARM Cortex-M, designed so that worst-case response times are analysed before they run, not benchmarked after, and privacy enforced below the application layer, where no application can bypass it.

Applications receive typed, consent-bound intent events — never raw neural streams.

It is not an AI-agent framework, a chatbot runtime or a token project. Every guarantee it makes is specified, openly licensed, and built to be checked by someone else.

Timing analysis that refuses rather than rounds. Worst-case response time for real-time systems, computed exactly in integer arithmetic — and a named refusal wherever a bound cannot be justified.

Two tasks. A runs 100 µs every 400 µs at higher priority; B runs 200 µs every 1,000 µs. The common answer for B is 400 µs. The correct one is 300 µs — and with other periods, the same mistake reports a deadline as met that is missed on hardware.

_{The schedule, drawn by simulating the scheduler rather than the formula. On 5,000 random task pairs it lands exactly on the analysis's answer.}

Try it live → · Source · The method · The bounty

Every figure published here, its evidence level, and the artefact it derives from. L1 formally proven · L2 measured on reference hardware · L3 independently reproduced · CI checked mechanically on every push. Graded as in CLAIMS.md, which this table follows. A figure absent from this table is not claimed.

No timing figure is claimed. A proof of a property, a proof of a time and a measurement are different kinds of statement, and only the first exists today. The first reference-hardware trace will land in axonos-validation, raw, whatever it shows; until then no microsecond figure appears in this table. L3 independent reproduction is not claimed for anything.

_{Not in this table, and therefore not claimed: classification accuracy, information transfer rate, power draw, on-hardware latency in a deployment, session length, electrode count in real use.}

One command, ninety seconds, no account:

git clone https://github.com/AxonOS-org/axonos-stack && cd axonos-stack
cargo run --locked --bin session -- --seed 7 --frames 3000 | diff - reference/session-7.txt

Silence means the whole chain — electrode to conditioning to privacy boundary to the right to act — reproduced byte for byte on your machine. The session is not a happy path: an electrode lifts partway through, and the transcript records the system withdrawing the right to actuate 96 ms later while it keeps recording.

Then the consent boundary, counted where the data arrives:

cargo run --locked --release --bin reference_bci | tail -1

RESULT: VERIFIED means that after consent is withdrawn at second 36, nothing — no raw sample, no derived reading, no decoded intent — reached the application, and that a second run reproduced every digest. How it works, and what it does not show →

# the full path, electrode to typed intent, verified bit for bit
git clone https://github.com/AxonOS-org/axonos-e2e-demo && cd axonos-e2e-demo && ./run.sh --verify

# the kernel: its tests, then a machine-checked proof
git clone https://github.com/AxonOS-org/axonos-kernel && cd axonos-kernel
cargo test --workspace
cargo kani setup && ( cd axonos-spsc/kani-proofs && cargo kani )

# the signal chain, bit-exact against conformance vectors
git clone https://github.com/AxonOS-org/axonos-signal-pipeline && cd axonos-signal-pipeline && cargo test

# the timing analysis, and every number it states against its source
git clone https://github.com/DYResearch/dy-wcet && cd dy-wcet && cargo test && ./audit.sh

A living map of every open-source brain–computer-interface project, tool and team, scored from public evidence and refreshed every three hours. AxonOS is ranked by the same formula as everyone else, with no boosting.

The State of Open BCI — read the full report →

_{Leading by reach: omi · wukong-robot · mne-python · NeuroKit · 15 languages · last refreshed 12 Oct 2026, 00:43 UTC}

The discipline behind AxonOS, applied to your system. DY Research carries out independent technical due diligence for investors, founders and engineering teams: every claim traced to its code, its tests and its evidence, ending in a written verdict.

Fixed scope, agreed in writing before any work begins. Revenue funds AxonOS. Engagements and full scope →

AxonOS does not currently claim, and this organisation must not be read as claiming: FDA clearance, CE marking or medical-device approval in any jurisdiction; clinical efficacy or independent clinical validation; certified medical-device status or production-implant readiness; complete compliance with IEC 62304, ISO 14971 or ISO 13485. These are possible future milestones, not present facts.

@article{axonos2026microkernel,
  title   = {An Analytical Microkernel Design for Safety-Critical
             Brain--Computer Interfaces: Schedulability, Capability
             Isolation, and Falsifiable Predictions},
  author  = {Yermakou, Denis},
  year    = {2026},
  doi     = {10.5281/zenodo.20552007},
  url     = {https://doi.org/10.5281/zenodo.20552007},
  note    = {Analytical bounds; predictions P1--P5; no measurement claims},
  license = {CC-BY-4.0}
}

Every repository exposes Cite this repository through CITATION.cff.