Become a sponsor to Project Navi
Project Navi builds open-source security infrastructure for teams deploying AI systems. MIT, APACHE 2.0, and AGPL licensed, zero dependencies where possible, deterministic guarantees where it matters. No venture capital, no corporate sponsors shaping the roadmap. The people who use the tools decide what matters. Your sponsorship keeps it that way.
What your support funds
[navi-sanitize] - Deterministic input sanitization for LLM pipelines. Homoglyphs, invisible Unicode, template injection, path traversal. Zero dependencies. Live on [PyPI].
[grippy-code-review] - Code review agent with a deterministic security rule engine and codebase knowledge graph. Structured findings, severity scores, pass/fail verdicts. Model-agnostic. Live on [PyPI].
[navi-bootstrap] - Spec-driven repo scaffolding. CI, security scanning, code review, and release pipelines in one command. 8 template packs. Live on [PyPI]
[navi-fractal] - Audit-grade fractal dimension estimation with refusal semantics. Four quality gates. Won't emit results without verified scaling regimes. Calibrated against formally proved networks.
[navi-SAD] - Spectral Attention Divergence. Dynamical systems probe for LLM hallucination detection via dual-path attention comparison and delay-coordinate attractor reconstruction.
Three Lean 4 formal verification repos back the toolchain - proving the mathematical foundations that navi-fractal and navi-SAD calibrate against. Zero sorry. CI-enforced.
Upstream security track record
CVSS 9.1 disclosure to OpenHands (merged to main). Responsible disclosure to Anthropic. SecurityAnalyzer suite merged to OpenHands software-agent-sdk (PR #2472). SimpleGraph.ball PR to Mathlib4 (in review). NIST RFI on AI agent security (docket NIST-2025-0035-0002).
Where every dollar goes
Compute, CI, dependency audits, documentation, and keeping these tools free. No marketing team, no sales org - just infrastructure that makes safe AI the path of least resistance.
The person behind this org is [@Fieldnote-Echo]. Sponsor them if you value the research and writing. Sponsor here if you depend on the infrastructure.
Machine cognition, human values.
Meet the team
-
Nelson Spence Fieldnote-EchoFounder & CEO. Seven years in behavioral health research, now building open-source AI security infrastructure.
Featured work
-
Project-Navi/navi-bootstrap
Spec-driven Jinja2 engine + 8 template packs for production-grade Python projects — CI, security, code review, and release pipelines.
Python 3 -
Project-Navi/grippy-code-review
AI-powered PR review agent with security rule engine, structured findings, and a grumpy auditor persona
Python 3 -
Project-Navi/navi-sanitize
Deterministic input sanitization for untrusted text — invisible characters, homoglyphs, and encoding tricks, handled before your code sees them. Zero dependencies, no ML. Python 3.12+.
Python 2 -
Project-Navi/navi-SAD
Spectral Attention Divergence — dynamical systems probe for LLM inference via dual-path attention comparison and delay-coordinate attractor reconstruction
Python 3 -
Project-Navi/navi-fractal
Sandbox fractal dimension estimation for complex networks — with quality gates, model selection, and refusal when evidence is insufficient
Python 2
0% towards $500 per month goal
Be the first to sponsor this goal!
$10 a month
Select- Sponsor badge. You believe formally verified security infrastructure should exist as a public good.
$25 a month
Select- Contributor. Name in SPONSORS.md across all repos.
$500 a month
Select- Organization. Logo on docs + quarterly call on verification priorities.
$2,000 a month
Select- Enterprise. Prominent placement + dedicated channel + annual strategy session + priority security advisories.