MacMind
A transformer neural network in HyperCard on a 1989 Macintosh
Details
- External ID
- 47792525
- Source
- HN
- Company
- —
- Product
- MacMind
- Website domain
- github.com
- Launched
- April 16, 2026
- Cohort
- —
- Upvotes
- 159
- Upvotes percentile
- 0.9395886889460154
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Description
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.
Enrichment
- Theme
- lightweight and on-device AI runtimes
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- Not AI
- Project type
- Hobby / open-source project
- Normalized one-liner
- neural network implementation in hypercard
- Manually corrected
- False
Could you build this?
No Implementing a transformer neural network with manual forward/backward autograd passes from scratch in HyperTalk on legacy 1989 Mac hardware requires extreme retrocomputing and low-level ML math expertise.
What it would actually take: The project requires mathematical derivation of multi-head self-attention, backpropagation, and Adam/SGD gradient descent implemented purely using HyperTalk scripting and card/stack state primitives. One must overcome extreme memory limits (system 6/7 Mac, few megabytes of RAM), integer/fixed-point numerical precision issues, and absence of matrix math libraries. This requires a unique convergence of deep vintage Apple Macintosh hardware hacking and fundamental mathematical machine learning.
Discussion
20 comments analyzed.
Competitors mentioned: PPC/SheepShaver, Infinite Mac, BasiliskII
Concerns raised: 32 KB script editor size limit due to TextEdit toolbox constraint, Interpreter speed performance challenges, Lack of arrays in HyperTalk, Resource fork preservation issues with modern macOS and Git
Feature requests: JIT compiler performance benchmarking, Support for larger script sizes without TextEdit limitation, Standalone compiler to write code separate from stacks
Competitors
Other products that read as similar to this one — 38 launches clear the similarity bar, closest 8 shown.
Attention rank: #6 of 39 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 150 days after the earliest competitor.
- I wrote my first neural network · hn · 2026-03-14 · 6 upvotes · similarity 0.41
- Duplicate 3 layers in a 24B LLM, logical deduction .22→.76. No training · hn · 2026-03-18 · 265 upvotes · similarity 0.39
- neural-network-digits · github · 2026-09-25 · 30 upvotes · similarity 0.38
- I built a lite LPU that can do inference on Karpathy's MicroGPT · hn · 2026-08-24 · 18 upvotes · similarity 0.38
- Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac · hn · 2026-07-29 · 919 upvotes · similarity 0.37
- LoRA gradients on Apple's Neural Engine at 2.8W · hn · 2026-03-06 · 6 upvotes · similarity 0.37
- RLT · github · 2026-09-13 · 51 upvotes · similarity 0.36
- recurrent-transformer · github · 2026-09-25 · 18 upvotes · similarity 0.36
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.