SyberLabs
empirical & experiential systems
Discover our work
the operators
Principles of Design
01

Stochastic

→ Unpredictability
noise · probability · surprise
02

Dynamic

→ Points in Time
attractors · feedback · regime shifts
03

Generative

→ Articulation
semantic fields · steering · decoder boundary
04

Geometric

→ Points in Space
symmetry · topology · proportion
05

Experiential

→ Subjectivity
audio · perception · presence
the laboratory

JEV integration is the team's number one technical priority across every project. The question that organizes the work: how may software change what people can perceive, understand, and create? SyberLabs is a laboratory for building and studying software at the intersection of machine learning, human-computer interaction, computational systems, simulation, and generative art. Work is attributed. Claims are graded. The systems are built to be entered, tested, and returned to.

Interfaces
Reading and thinking surfaces: RISE, OmniOS.
Kernels
Typed hypergraph rewrite with exact stochastic semantics: OSAHR.
Evidence
Public claims stay labeled. Frozen artifacts carry checksums.
the people
Who builds it

Mateo Robles

Principal Architect

Engineer. Sets the form of the work and builds a large part of it: what a system is for, how it should feel, and the product that ships.

  • Product design and UI/UX for RISE and OmniOS: the Chamber, the canvas, the surfaces a person actually meets.
  • Engineering in the same hand: client runtimes, typed ports, data and key paths, the OSAHR experiment line and public 0.2 freeze.
  • Vision for the laboratory: what the systems are, and what they must not pretend to be.
VisionProductUI/UXEngineering

Seth Carlson

Principal Engineer

Engineer. Measures the system, deletes what should not exist, and puts production under control.

  • Relay: a job-search review workspace still in early release. Research history, exact draft acceptance, application follow-ups.
  • RISE: stopped shipping books as JavaScript, so first load is the app, not the library.
  • RISE: production control. A first-load budget that fails a broken deploy, CI from ten minutes to five, architecture generated from the code.
  • Discipline against feature creep: delete first, then refuse a second engine for a job that already has a path. Same cut on OSAHR unused kernel surface.
BundleProductionDisciplineKernel
selected work
Built & Tested

JEV integration

Team-wide top priority

JEV integration is SyberLabs' number one technical priority across every project. The team is actively recruiting engineers with hands-on JEV experience.

  • JEV integration is the shared technical focus across the portfolio.
  • We are actively recruiting engineers who have built, integrated, or operated JEV systems.
Experienced with JEV? Contact the team ↗
JEVIntegrationDecision routingHiring

Relay

Application / review

A job-search review workspace: one history for repeated postings, source records, and exact drafts. Assistants hand off; a person accepts.

  • Consolidate posting URLs and keep research history. Notes and interview follow-ups edit without resetting status.
  • Accept an exact draft; changing it returns it to review. Generated claims still require a person.
  • Early release. Local handoffs, not a finished product. No background sync, no autonomous applying.
TypeScriptNotionClaudeReview

RISE

Interface

A browser-based environment for reading text through time, image, sound, and procedural form. One source; two projections; files never leave the browser.

  • The Chamber is one instrument: text, time, image, and sound compile into Stream (time) or Page (space).
  • Local files stay in the browser. Absence is preferred to false substitution; provenance travels with the work.
  • Client-only Vite app in open beta: a reading surface another person can enter and return to.
ReadingAudioCanvasWeb

OmniOS

Product / HCI

A spatial canvas for thinking with AI over live data: blocks, wires, and personas whose context is what the wires actually carry.

  • Typed ports; a persona's context is inspectable; provenance comes from the turn, not from empty connections.
  • Keys stay server-side. Public sources run without keys; everything else falls back to built-in mock data.
  • Local-first and single-user. The canvas is the product.
TypeScriptCanvasHCIAI

OSAHR

Systems / Kernels

A correctness-first kernel for stochastic adaptive rewrite over a typed directed hypergraph. Exact schedulers. Claim licenses. No silent approximate shortcuts.

  • Injective typed DPO matches; a dangling-invalid match is not assigned a rate and cannot steal probability.
  • Exact SSA, modified next-reaction, and bounded thinning; incremental matching is checked against an exhaustive oracle.
  • Experiments grade claims KNOWN, MEASURED, INFERRED, or PROPOSED. Artifacts freeze with checksums.
PythonHypergraphsSimulationEvidence

We are actively recruiting engineers experienced with JEV. If you have built, integrated, or operated JEV systems, contact the team.

contact github