MacMind – A transformer neural network in HyperCard on a 1989 Macintosh
I trained a transformer in HyperCard. 1,216 parameters. 1989 Macintosh. And yes, it took a while. MacMind is a complete transformer neural network, embeddings, positional encoding, self-attention, backpropagation, and gradient descent, implemented entirely in HyperTalk, the scripting language Apple shipped with HyperCard in 1987. Every line of code is readable inside HyperCard's script editor. Option-click any button and read the actual math. The task: learn the bit-reversal permutation, the opening step of the Fast Fourier Transform. The model has no formula to follow. It discovers the…
In plain words
MacMind is a complete transformer neural network built in HyperTalk, the scripting language for Apple's 1987 HyperCard software, running on a 1989 Macintosh. It includes embeddings, positional encoding, self-attention, backpropagation, and gradient descent—all readable and editable directly in HyperCard's script editor. The 1,216-parameter model learns to recognize bit-reversal permutations, a step in the Fast Fourier Transform, discovering patterns purely through attention and repeated training cycles without explicit formulas.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
I trained a transformer in HyperCard. 1,216 parameters. 1989 Macintosh. And yes, it took a while. MacMind is a complete transformer neural network, embeddings, positional encoding, self-attention, backpropagation, and gradient descent, implemented entirely in HyperTalk, the scripting language Apple shipped with HyperCard in 1987. Every line of code is readable inside HyperCard's script editor. Option-click any button and read the actual math. The task: learn the bit-reversal permutation, the opening step of the Fast Fourier Transform. The model has no formula to follow. It discovers the positional pattern purely through attention and repeated trial and error. By training step 193, it was oscillating between 50%, 75%, and 100% accuracy on successive steps, settling into convergence like a ball rolling into a bowl. The whole "intelligence" is 1,216 numbers stored in hidden fields in a HyperCard stack. Save the file, quit, reopen: the trained model is still there, still correct. It runs on anything from System 7 through Mac OS 9. As a former physics student, and the FFT is an old friend, it sits at the heart of signal processing, quantum mechanics, and wave analysis. I built this because we're at a moment where AI affects all of us but most of us don't understand what it actually does. Backpropagation and attention are math, not magic. And math doesn't care whether it's running on a TPU cluster or a 68030 from 1989. The repo has a pre-trained stack (step 1,000), a blank stack you can train yourself, and a Python/NumPy reference implementation that validates the math.
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 · 27d ago · cactuscompute.com
Launched alongside, April 2026
the whole month →
- AG
Thought the resources for GPU arch were lacking, so here we are
Life & fun · Apr 2026 · jaso1024.com
- IB
Built a ~9M param LLM from scratch to understand how they actually work. Vanilla transformer, 60K synthetic conversations, ~130 lines of PyTorch. Trains in 5 min on a free Colab T4. The fish thinks the meaning of life is food. Fork it and swap the personality for your own character.
AI · Apr 2026 · github.com
- BC
Life & fun · Apr 2026 · sam-burns.com
- IB
With social media and now AI, its important to keep the indie web alive. There are many people who write frequently. Blogosphere tries to highlight them by fetching the recent posts from personal blogs across many categories. There are two versions: Minimal (HN-inspired, fast, static): https://text.blogosphere.app/ Non-minimal: https://blogosphere.app/ If you don't find your blog (or your favorite ones), please add them. I will review and approve it.
AI · Apr 2026 · text.blogosphere.app