First screen, captured 3 days ago. Hover to scroll the page, or open the video.

Show HN: A working Nintendo in ~1GB of Qwen weights

What HN thinks

A Nintendo running 200x slower inside an LLM: delightful, horrifying, and a reminder that plain code still wins at being code.

Fun hack, shows LLM inefficiency
Many enjoyed it but stressed that running at 150–290x slower than real hardware shows how vastly more efficient ordinary deterministic programs are than computing through weights.
Good specs and tests enabled it
The author and others credit the emulation community's conformance tests and the Qwenputer compiler for letting Opus build it in two prompts, and note that test-driven specs like Gherkin work well with LLMs.
Pointless slop or old idea
Skeptics said reducing logic circuits to weights is trivial or derivative of Neural Turing Machines, and complained about the AI-written README.

Consensus: Most agree it is a clever, fun demonstration that is not practical, and that real emulators and code remain far more efficient.

Recommended in the comments

Gherkin ↗ (tool): Commenters said feeding Claude Gherkin-formatted test criteria yields very good output.
Qwenputer ↗ (library): It is the foundation this project builds on and commenters found it very cool.
rhun ↗ (product): Recommended as an example of AI-written assembly software that is genuinely useful daily.

Summarised by Claude from the comment thread 2 days ago.

Points: 278
Comments: 136
Best rank: #1
Posted: 2026-10-30 by dbreunig
Category: Ai ml

nes-emulation qwen transformer-compilation 6502 gated-deltanet python

All captures:

Full page

Captured 3 days ago. Click for the whole screenshot.

Looks similar

Screenshot of github.com: Show HN: I Turned DOOM into a Transformer

Show HN: I Turned DOOM into a Transformer (github.com)

120 points by falperc 3 days ago | 35 comments

GitHub repository for DOOMFORMER, a compiler that turns an integer neural network into a DOOM map where voodoo dolls perform the arithmetic, verified against NumPy.

doom transformer neural-networks esoteric-computing

Screenshot of github.com: Show HN: OmniDiff – Opensource, super fast, content aware Git difftool, in Rust

Show HN: OmniDiff – Opensource, super fast, content aware Git difftool, in Rust (github.com)

294 points by FabCH 1 day ago | 91 comments

GitHub repository for OmniDiff, an open-source AGPL Rust git difftool that does content-aware diffing of code, configs, images, PDFs, archives and fonts.

git difftool rust structural-diff

Screenshot of github.com: Show HN: Lasm – Write formally verified webapps in Lean for Node/serverless/edge

Show HN: Lasm – Write formally verified webapps in Lean for Node/serverless/edge (github.com)

25 points by lastmjs 3 days ago | 4 comments

GitHub repository for Lasm, a toolchain for compiling Lean 4 programs to JavaScript and WebAssembly runtimes such as Node, browsers, serverless and edge platforms.

lean4 formal-verification webassembly serverless

Screenshot of github.com: Show HN: Ghostman – A fast, keyboard-driven TUI API client and webhook inspector

Show HN: Ghostman – A fast, keyboard-driven TUI API client and webhook inspector (github.com)

49 points by 4yk5 3 days ago | 26 comments

GitHub repository for Ghostman, a Python/Textual terminal API client and local webhook inspector positioned as a local-first Postman alternative.

tui api-client webhooks python

Screenshot of github.com: Astroterm: A Planetarium for Your Terminal

Astroterm: A Planetarium for Your Terminal (github.com)

213 points by nvmdbljstm 5 days ago | 25 comments

GitHub repository for astroterm, an MIT-licensed C program that renders stars, planets, and constellations as a planetarium in the terminal.

astroterm terminal tui astronomy

Screenshot of github.com: Show HN: Live speech-to-text on an ESP32-P4 microcontroller, no cloud

Show HN: Live speech-to-text on an ESP32-P4 microcontroller, no cloud (github.com)

9 points by boumedinebillal 6 days ago | 3 comments

GitHub repo for esp32-pie-asr, a 2/4-bit quantized 65.7M-parameter streaming speech recognition model running live on an ESP32-P4 microcontroller without cloud.

speech-recognition esp32-p4 tinyml quantization

Full page

Looks similar

Show HN: I Turned DOOM into a Transformer (github.com)

Show HN: OmniDiff – Opensource, super fast, content aware Git difftool, in Rust (github.com)

Show HN: Lasm – Write formally verified webapps in Lean for Node/serverless/edge (github.com)

Show HN: Ghostman – A fast, keyboard-driven TUI API client and webhook inspector (github.com)

Astroterm: A Planetarium for Your Terminal (github.com)

Show HN: Live speech-to-text on an ESP32-P4 microcontroller, no cloud (github.com)

Related stories

Show HN: I rebuilt Kubernetes for ESP32 microcontrollers (github.com)

Show HN: Witr – Explain why a process is running on your Linux system (github.com)

CuPy: NumPy and SciPy for GPU (github.com)

qm – Multiplayer agent harness for work (github.com)

Show HN: Wat – Deep inspection of Python objects (github.com)

FSRS: A modern, efficient spaced repetition algorithm (github.com)