AI with intelligent guardrails.

Where knowledge is forged, not forgotten. Thoughtful software for everyday life and enterprise.

Research Update: We Tested Our Retrieval Thesis. It Failed. Here’s What Survived.

A different kind of AI company.

Most AI companies optimize for engagement, growth, and data extraction. We optimize for helpfulness, simplicity, and trust.

Our first rule is "do no harm."

Three pillars of innovation

Consumer Applications

Simple, ethical apps for everyday life. No dark patterns. No exploitation. Just helpful tools that respect your time.

Learn more

Enterprise AI

Intelligent systems for business workflows. Reduce costs. Increase capability. Maintain trust.

Learn more

Research

Advancing AI memory and reasoning. Rigorous research. Practical applications.

Learn more
SpotlightRESEARCH

ngramDB

A memory substrate under measurement. We test our own claims with pre-registered experiments—and publish what the benchmarks say, including when they say no.

Tested, Not Asserted

We built a biologically inspired memory system, then ran it against strong baselines under pre-registered pass/fail criteria. The retrieval hypothesis failed its test—so we closed it and said so. What survived is a substrate for memory you can audit.

  • Durable, replayable memory storage
  • Typed causal extraction with full provenance
  • Pre-registered, published kill tests
  • Persona isolation for multi-context use

Retrieval thesis: closed

Graph vs. similarity, measured head-to-head. Similarity won; we published it

Substrate integrity: kept

Bit-identical replay, WAL durability, crash recovery, full provenance

Now testing: consequence at encoding

Causal annotations inside the representation—pre-registered, publish either way

In design: memory that acts

Wake conditions and evidence-carrying answers, gated on validation

Let's work together

Interested in our enterprise solutions or want to learn more?

Get in Touch