Bootloops

Bootloops is a scientific software package and working protocols for large language models doing precision quantitative science.

BootLoops 1.0 is an extensive scientific software package written and upgraded to be driven by a large language model, along with working protocols that make results checkable. The system is independent of the model driving it, giving agents instruments to run with documented acceptance gates, outputs, and verification tests. It handles frontier integrals of mathematical physics, recurrences with proofs, Bayesian evidence integrals, ball arithmetic for effective error propagation, and exhaustive enumeration with completeness certificates. The toolkit includes packages such as JaCKandJill for Bayesian phylogenetics, Mixalot for statistics, Terrier, and Popcorn for population genetics. It also provides house engines like SOFIA.jl, Eichler.jl, and the Leviathan Landau engine, alongside patched forks of Kira, Blade, and AMFlow.cpp.

Bootloops at a glance

Best for: Software engineer, Researcher, Student
Runs on: Linux, API, CLI, self-hosted, open source
Pricing: Free
Released: October 2026
Category: AI research tool
Website: github.com/bootloops-ai/bootloops

Key features

Frontier integrals of mathematical physics
Recurrences with proofs and certificates
Bayesian evidence integrals in closed form
Ball arithmetic for effective error propagation
Exhaustive enumeration with completeness certificates
Open implementations of standard statistical procedures
House engines for exact physics and quantitative science

Use it for

Check published numbers and integrals from papers
Compute multi-dimensional parametric integrals
Run Bayesian evidence calculations for phylogenetic models
Execute certified error bounds instead of statistical sampling

Alternatives to Bootloops

Hypoarena — Provides a fully offline environment to test scientific discovery pipeline logic.
Vegaml — Runs locally on your hardware with calibrated probabilities and conformal answer sets.
OpenResearch — Run parallel research workflows locally or on remote compute.
FAROS — Connects literature evidence, executable plans, and human review into an auditable research loop.
Math — Contains 722 mathematical manuscripts organized into 372 families.
Open Science — Local-first design with persistent Python and R kernels and traceable artifacts.
Open Math OS — It unites open frontier models, scientific compute, and data sovereignty in one workspace.
AI Scientist v3 — Automates end-to-end scientific research from task descriptions to complete papers.
SemIf OpenJev — Extracts direct token probabilities without generating text or running JSON repair loops.
Lemma — Read mathematics with interactive proofs, reference links, and checked derivations.
Finetuning With Sampling — Official implementation of the Finetuning with Sampling research paper from Harvard.
Auto Deep Researcher — Automates experiment operations to reduce manual babysitting of training loops.

All Bootloops alternatives