SYBERLABS
SyberLabs / Independent AI software lab
Read.
Think.
Build.
SyberLabs builds AI-powered reading software and tools that make AI agents reviewable.
Work
Projects
- 01RISEAI-assisted reading appRead beyond the page.A browser-based reading app that controls the pacing, layout, imagery, and sound around a text, with AI-suggested settings.JavaScript, Vite, Cloudflare Workers, PostgreSQL (Neon), Redis (Upstash), sql.js, Vitest, PlaywrightLive
- 02CommonsProject coordinationTurn a need into a mission.A prototype for community-led projects that keeps the need, plan, evidence, human review, and outcome in one auditable record.Private prototype
- 03RelayJob application workspaceEvery application has a history.A workspace for job seekers who draft with AI assistants, with one record per job and human approval of the final wording.TypeScript, React, Cloudflare Workers, Drizzle ORM, Chrome extension (Manifest V3), PlaywrightEarly release
- 04OmniOSAI analysis canvasSee the sources behind an answer.A canvas for asking AI questions about live data, where every answer shows which sources it used.TypeScript, Next.js, React, Zustand, IndexedDB, PostgreSQL, Tailwind CSS, Vitest, PlaywrightResearch
- 05OSAHRStochastic simulation libraryReplay a changing system.An open-source Python library for exact stochastic simulation of networks that change their own structure, with step-by-step replay.Python 3.11+ (no runtime dependencies in the core library), pytestResearch
Research / September 2026
Reliable execution for AI agents.
A model can choose an action, but an application still has to enforce permissions and budgets, check the result, and recover from crashes. Our current research defines that execution layer and tests it with a Python and SQLite prototype that passes 87 regression tests, including process-crash recovery experiments.
87passing regression tests
Read the technical reportAbout
About SyberLabs
SyberLabs is an independent software and AI research lab founded in 2026 by Mateo Robles. The lab designs, builds, and evaluates AI products and research software, with a focus on applied machine learning, LLM applications, AI agent reliability, and simulation.
Every project is published with its current status and its limits. Prototypes, measured results, and planned work are labeled separately so readers can tell what has been shown and what has not.
Technical skills
- Python
- TypeScript
- JavaScript
- SQL
- React
- Next.js
- FastAPI
- PostgreSQL
- SQLite
- Redis
- Cloudflare Workers
- LightGBM
- scikit-learn
- pandas
- NumPy
- LLM applications
- AI agents
- pytest
- Vitest
- Playwright
