SymDerive – A functional, stateless symbolic math library
Hey HN, I’m a physicist turned quant. Some friends and I 'built' SymDerive because we wanted a symbolic math library that was "Agent-Native" by design, but still a practical tool for humans. It boils down to two main goals: 1. Agent Reliability: I’ve found that AI agents write much more reliable code when they stick to stateless, functional pipelines (Lisp-style). It keeps them from hallucinating state changes or getting lost in long procedural scripts. I wanted a library that enforces that "Input -> Transform -> Output" flow by default. 2. Easing the transition to Python: For many…
In plain words
SymDerive is a symbolic math library designed for AI agents and physicists transitioning to Python. It uses stateless, functional programming patterns to improve agent reliability and reduce hallucinations in AI-generated code. The library wraps standard Python scientific tools while maintaining familiar Mathematica-style syntax, providing a bridge between mathematical tradition and modern Python workflows.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hey HN, I’m a physicist turned quant. Some friends and I 'built' SymDerive because we wanted a symbolic math library that was "Agent-Native" by design, but still a practical tool for humans. It boils down to two main goals: 1. Agent Reliability: I’ve found that AI agents write much more reliable code when they stick to stateless, functional pipelines (Lisp-style). It keeps them from hallucinating state changes or getting lost in long procedural scripts. I wanted a library that enforces that "Input -> Transform -> Output" flow by default. 2. Easing the transition to Python: For many physicists, Mathematica is the native tongue. I wanted a way to ease that transition—providing a bridge that keeps the familiar syntax (CamelCase, Sin, Integrate) while strictly using the Python scientific stack under the hood. What I built: It’s a functional wrapper around the standard stack (SymPy, PySR, CVXPY) that works as a standalone engine for anyone—human or agent—who prefers a pipe-based workflow. # The "Pipe" approach (Cleaner for agents, readable for humans) result = ( Pipe((x + 1)**3) .then(Expand) .then(Simplify) .value ) The "Vibes" features: Wolfram Syntax: Integrate, Det, Solve. If you know the math, you know the API. Modular: The heavy stuff (Symbolic Regression, Convex Optimization) are optional installs ([regression], [optimize]). It won’t bloat your venv unless you ask it to. Physics stuff: I added tools I actually use—abstract index notation for GR, Kramers-Kronig for causal models, etc. It’s definitely opinionated, but if you’re building agents to do rigorous math, or just want a familiar functional interface for your own research, this might help. I have found that orchestrators (Claude Code, etc) are fairly good at learning the tools and sending tasks to the right persona, we have been surprised by how well it has worked. Repo here: https://github.com/closedform/deriver I will cry if roasted too hard
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
