AI is fluent.
It is not grounded.

Language models write confident answers that can be wrong — and cannot reliably say why they said what they said. In research, clinical work, and tools for vulnerable users, an eloquent guess is a liability. Bigger models and bolt-on retrieval make the guesses smoother, not the grounding real.

Let the symbols
author the facts.

Peel flips the usual architecture. A typed symbolic store is the only author of facts. The neural model is demoted to two jobs — propose (it must pass symbolic extraction) and render (it cannot add content). Hallucination is prevented structurally, not statistically.

One console.
Every answer sourced.

• Peel — a typed knowledge graph where every fact carries its provenance.
• Lois — a reasoning core that answers supported, conflicting, or unknown.
• The Perslis runtime — 44 source-pinned research tools on one console, driven in plain language.

Get your lab hooked up.
No special equipment.

One console connects the systems a lab already uses — 19 public science services (PubMed, UniProt, AlphaFold, ClinVar, KEGG and more), a compute bench, and WetHands lab automation reaching LIMS, electronic notebooks, and instrument APIs. We connect the instruments that were never meant to connect. Nothing to rip out.

A contract nothing
gets past.

• Source contract — a result that cannot name its source is refused, by construction.
• Air-gap ready — the lab record is local files that keep working with the cable pulled.
• Metal to wet bench — pure C++ on the metal and a wet-lab lane on the same console.
• A symbolic admission layer — not trained guardrails that can be talked around.

Your research never
leaves your computer.

• Local models by default. Ollama runs the reasoning on your own machine — no prompt, file, or graph is sent to a cloud model to answer a question.
• We never train on your data. Nothing leaves your devices to train any model, ever. Your research is not someone else’s training set.
• External models are opt-in. Want a hosted model for a task? Choose from many providers — only if and when you decide. Off by default, always your call.
• The record is yours. Graph, ledger, notebook, and files live on your disk and keep working with the cable pulled.

Adoption is here.
Trust is the bottleneck.

AI is entering regulated and high-stakes work faster than it can be trusted there. Agentic tools now act, not just chat — so access has to be governed and answers have to be checkable. On-prem and air-gapped demand is real. Grounding is the gate everyone hits, and it is the gate Peel is built to be.

Research runs on
tools like these.

• Bioinformatics & research tooling — the teams that live inside these databases every day.
• Assistive & AAC — where Peel began: context and dependable expression for vulnerable users.
• Regulated & enterprise research — pharma, clinical, and government, where provenance is non-negotiable.

Land a pilot.
Expand the lab.

We stand up a pilot lab with a research partner around one bounded question and a clear definition of success. Pilots convert to per-server and per-seat licensing; enterprise deploys on-prem. Integration and support compound the relationship. Land narrow, expand across the lab.

Prototype to
enterprise.

Fund the runtime
research can trust.

We are raising to productize the runtime for on-prem and enterprise, validate it with research partners against expert references, and build go-to-market. The vision: a research environment where every conclusion carries its evidence — and no answer outruns what can be shown.