Back to browse
OtherFunc – Serverless functions in Brainfuck, Forth, BASIC, and more

OtherFunc – Serverless functions in Brainfuck, Forth, BASIC, and more

by mdeeds·Feb 18, 2026·4 points·0 comments

AI Analysis

●●SolidRabbit HoleCozy

Brainfuck can now make HTTP requests; technically impressive, practically a novelty.

Strengths
  • Memory-mapped I/O hack in Brainfuck (~35k char program to do HTTP) shows real ingenuity—turns Turing-tarpit language into something remotely functional.
  • Multi-language interpreter compilation to single Wasm binary deployed on Cloudflare Workers is elegant architecture, zero-dependency deployment.
  • Comprehensive feature matrix (KV storage, env vars, HTTP) across diverse interpreter semantics shows thoughtful platform design for a joke category.
Weaknesses
  • Esoteric languages have no production user base; the appeal is educational/entertainment, not a solved business problem or unmet developer need.
  • Execution caps (500K instructions) and 100KB code limits make it unsuitable for real workloads despite author's 'production if feeling adventurous' framing.
Target Audience

Esolangs enthusiasts, educational users, novelty seekers, developers wanting unconventional infrastructure experiments

Similar To

Replit · TIO.run

Post Description

Hi HN, This started as a weekend brainfuck interpreter and kept growing. I got curious whether you could make a usable serverless platform out of languages that were never intended for this.

OtherFunc is a serverless function platform for languages no major cloud provider (to my knowledge) supports. There are currently implementations of brainfuck, Forth, APL, Lisp (Scheme-like), and BASIC.

The interpreters are written in Rust, compiled to a single Wasm binary, and deployed on Cloudflare Workers. You can finally write Forth and publish it as an HTTP endpoint. You can see some examples on the showcase page: https://otherfunc.com/showcase

- brainfuck can make HTTP requests. The tape is extended to 33,000 cells with a memory-mapped I/O region. You write a URL to cells 30,000+, set a method byte, trigger execution, and the response appears in cells 31,000+. The Brainfuck program to do this runs ~35,000 characters, but it works.

- The interpreters use a coroutine/yield pattern for I/O instead of async. When code needs to make an HTTP call or access KV storage, the interpreter suspends with an IoRequest, the Worker performs the fetch, then resumes execution with the response.

- There's an MCP server so AI assistants can deploy functions directly. The thought was, if an LLM is writing all your code anyway, the language it is written in doesn't really matter. But you probably don't want to waste your tokens writing bf either.

Code is available here: https://github.com/otherfunc

Similar Projects