Skip to content
HN On Hacker News ↗

MathCode — A Frontier Mathematical Coding Agent

▲ 91 points 28 comments by homarp 16h ago HN discussion ↗

Pangram verdict · v3.3

We believe that this text is a mix of AI and human-written content.

67 %

AI likelihood · overall

Mixed
28% human-written 72% AI-generated
SEGMENTS · HUMAN 1 of 4
SEGMENTS · AI 1 of 4
WORD COUNT 301
PEAK AI % 86% · §3
Analyzed
Aug 16
backend: pangram/v3.3
Segments scanned
4 windows
avg 75 words each
Distribution
28 / 72%
human / AI fraction
Verdict
Mixed
Pangram v3.3

Article text · 301 words · 4 segments analyzed

Human AI-generated
§1 Mixed · 63%

Overview MathCode is a terminal AI coding assistant with a built-in math formalization engine. Give it a math problem in plain language and it will automatically convert it into a Lean 4 theorem and attempt a formal proof — with a persistent Lean REPL, reusable theorem and axiom libraries, agentic proving, and an Obsidian knowledge graph. Quick Start Requires macOS (arm64) or Linux (x86_64), plus the codex CLI for the default backend.

§2 Mixed · 56%

git clone https://github.com/math-ai-org/mathcode.git cd mathcode bash setup.sh codex auth login mathcode setup.sh prepares the release checkout, downloads the bundled runtime and Lean toolchain, and installs a user-local mathcode launcher.

§3 AI · 86%

Try it with: mathcode -p "prove that the square of an even number is even" Outputs are written to LeanFormalizations/. A browser UI is available via ./run webui. Features Persistent Lean REPLA persistent Lean language server brings compile checks to ~0.4s after a one-time warmup, instead of ~30s. Theorem LibraryEvery proved theorem is auto-named, stored, and made importable so the prover and planner can reuse it. Axiom LibraryStore conversational assumptions as persistent, compile-checked, consistency-reviewed Lean declarations. Lean LSP IntegrationSearches leansearch.net and Loogle for verified Mathlib lemmas and uses structured LSP diagnostics for repairs. Obsidian Theorem GraphGenerates an Obsidian vault that visualizes theorem-to-lemma dependencies as a knowledge graph. Agent-Mode ProvingEach proof becomes an interactive session where the agent writes candidates, reads errors, and recompiles. Tree-of-SubgoalsDecomposes complex theorems into independent subgoals and proves them in parallel, then stitches them back. Multi-PlannerRuns multiple planners in parallel for diverse proof strategies; the prover picks the best approach.

§4 Human · 20%

Citation If you use MathCode in research, please cite: @misc{mathcode2026, title = {MathCode: A Frontier Mathematical Coding Agent}, author = {Team Math-AI}, journal = {math-ai-org.github.io}, year = {2026}, month = {April}, url = {https://github.com/math-ai-org/mathcode} } The math formalization and proving pipeline is based on the AUTOLEAN project.