Start from zero.
Build what lasts.

Zeroline Labs is an independent research laboratory studying the foundations of intelligent systems, and building technology from first principles.

The defining technologies of this century will be built by those willing to begin at the foundations.

We work on hard, foundational problems in computing and intelligence, where careful definitions, rigorous reasoning and honest measurement matter more than speed.

We publish what we can prove and release what we can stand behind. Every result is measured against the same fixed reference: the line we draw before the work begins.

Four questions guide our work.

Our research is organised around open questions. Each is fundamental, measurable, and worth years of careful work.

Fig. 1 — Convergence to the reference liney(t) = A·e−kt·sin(ωt + φ) → 0, for 24 initial conditions
  1. 01Foundations

    What makes learning generalise?

    The mathematics of representation, optimisation and generalisation: why learning systems work, and when they will fail.

  2. 02Verification

    How can we prove a system does what it should?

    Formal specification, program logics and verified software for systems where testing alone is not enough.

  3. 03Efficiency

    How little computation does intelligence require?

    Algorithms and architectures that treat compute, data and energy as first-class constraints, not afterthoughts.

  4. 04Interaction

    How should people and intelligent systems work together?

    Tools and interfaces that extend human judgement, keep people in control, and earn their trust.

Latest work

All research
Forthcoming

Our first papers and research notes are in preparation.

Each will be published here once its results have been completed, checked and reproduced. To be notified when the first appears, write to hello@zerolinelabs.com.

From zero, in four steps.

  1. 0

    Define

    State the problem precisely. Assume nothing that has not been shown.

  2. 1

    Derive

    Reason from first principles to a design that can be explained.

  3. 2

    Build

    Implement with care, at the smallest scale that answers the question.

  4. 3

    Measure

    Test against the reference set at the start. Keep only what holds.

What we hold ourselves to.

  1. I

    First principles

    “What I cannot create, I do not understand.”
    Richard Feynman
  2. II

    Proof over opinion

    “Program testing can be used to show the presence of bugs, but never to show their absence.”
    Edsger W. Dijkstra, 1970
  3. III

    Simplicity as a requirement

    “Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away.”
    Antoine de Saint-Exupéry, 1939
  4. IV

    Elegance as evidence

    “Beauty is the first test: there is no permanent place in the world for ugly mathematics.”
    G. H. Hardy, 1940
  5. V

    Built to last

    “Firmitas, utilitas, venustas. Strength, utility, beauty.”
    Vitruvius, De architectura

Software, carefully made.

Tools for everyday work. Each does one thing precisely, and is maintained for as long as people rely on it.

All products

A small group, deliberately.

We are forming a small team of researchers and engineers with deep foundations in mathematics, computer science and engineering.

We value clear thinking, rigorous writing and patient craftsmanship. If that describes how you work, we would like to hear from you.