Skip to content
PerslisMachine Minds
Try Perslis

AI IS AMAZING AT

Sounding right.Even when it's making stuff up.

Meet Perslis. We build machine minds that check the facts, follow explicit rules, and show their work. Because confidence is great. Evidence is better.

Try a live example
A dog in a Perslis hoodie beside a tangled robot covered in uncertain answers

Rules, not hidden weights.

Local, even at the edge.

Every decision inspectable.

Real-world readyFrom legacy to next-gen

The symbolic company

Intelligence
without the
hallucinations.

Large Language Models can invent facts, exposing your company to operational, legal, and reputational risk. Perslis treats a 15% failure rate as unacceptable. Learned rules are checked, reviewed, and authorized before they can act—decisions run on evidence, not vibes.

Perslis Machine Minds: evidence, authorization, and action in one inspectable loop.

No black boxes

See the
reasoning.

Every decision is logged, inspectable, and reproducible. Explore the rules, evidence, and execution trace yourself.

Try it live View example traces →
User request

Migrate this COBOL report to a modern schema and validate the results.

Interactive sample · runs locally.
Click a step or tab to inspect it.

Loading local sample…

    One engine. Many worlds.

    Real problems.
    Real impact.

    From legacy systems to scientific discovery, from accessibility to autonomy—Perslis brings symbolic, verifiable reasoning to the challenges that matter. Explore all use cases →

    Legacy systems

    Operate what cannot be replaced.

    Understand and control the software your organization still depends on.

    See the use case
    Scientific discovery

    Trace every result to evidence.

    Connect research sources, tools, reasoning, and inspectable records.

    See the use case
    Accessible communication

    Turn intent into a voice.

    Support communication for children, adults, and stroke survivors.

    See the use case
    Bounded defense Autonomy research

    Adapt the plan.
    Keep authority fixed.

    See the use case

    Perslis proof

    See it work

    Open the demos. Explore the decisions.

    Legacy application demonstration▶Demo

    Operate a legacy applicationLive proof
    Perslis Science research demonstration▶Demo

    Trace a research resultLive workspace
    Perslis Motion driving demonstration▶Demo

    Watch a motion rule in actionSimulation
    Layered mountain ridges in morning mist

    A brighter tomorrow

    Machine minds
    for a safer world.

    Safer systems. Greater access. Real-world impact.
    Let’s turn what’s possible—into what’s real.

    Get started Talk to our team

    Perslis models

    Introducing Lois

    Perslis’s symbolic reasoning model for scientific research. Lois connects facts to sources and exposes the rules and evidence behind an answer. Missing evidence stays visible.

    MODEL
    Lois
    CATEGORY
    Scientific reasoning
    PRINCIPLE
    No source, no claim
    Read the research paper →

    Introducing FailFirst

    A model that makes failure useful without making authority negotiable. Observed failures become readable rules that can narrow behavior, never widen it.

    MODEL
    FailFirst
    CATEGORY
    Bounded learning
    PRINCIPLE
    Fail closed
    Read the research paper →

    Introducing Symbols

    A hybrid model built to convert an existing system into symbolic traces. It connects learned interpretation to explicit operations, rules, and inspectable outputs.

    MODEL
    Symbols
    CATEGORY
    Hybrid symbolic tracing
    PRINCIPLE
    Turn systems into traces
    Read the research paper →

    Introducing TinkyMind

    A small, local model for communication that keeps working with the cable out. It runs on-device, offline, without sending the conversation to the cloud.

    MODEL
    TinkyMind
    CATEGORY
    On-device language
    PRINCIPLE
    Nothing on the wire
    Read the research paper →