# tinygrad

40 Hacker News commenters recommended tinygrad and 16 criticized it, November 2025 to November 2026. It was recommended in 9 of 13 months. It meets the Hacker Approved badge rule.

**Verdict:** Praised for a small, readable deep learning framework that competes with PyTorch in far less code and reaches many hardware backends. Common caveat: little production use yet and slow inference.

## Why people recommend it

- Small, readable codebase (5 comments): "There isn't any part of it that looks terribly hard to understand when compared to other code." ([source](https://news.ycombinator.com/item?id=46441294)) "I don't think this is bad, if you know Python then most of the code will be fine for you." ([source](https://news.ycombinator.com/item?id=46442607)) "A 1000x reduction in code size and matching/exceeding in perf is worth looking at." ([source](https://news.ycombinator.com/item?id=46438108))
- Runs on many hardware backends (4 comments): "tinygrad does have pretty neat support for sharding things across various devices relatively easy" ([source](https://news.ycombinator.com/item?id=47641540)) "the same stuff allows direct gpu communication so that even mac nvidia thing is possible" ([source](https://news.ycombinator.com/item?id=46380267)) "He was able to run nvidia gpu on mac via thunderbolt with tinygrad." ([source](https://news.ycombinator.com/item?id=46424916))
- Direct path to hardware (3 comments): "the most interesting thing about tinygrad was that theoretically you could render a model all the way into hardware" ([source](https://news.ycombinator.com/item?id=47471966)) "The TinyGrad approach of going straight to the hardware is telling" ([source](https://news.ycombinator.com/item?id=47756897)) "the push for a more vertical integration with tinygrad" ([source](https://news.ycombinator.com/item?id=47476039))
- Competes with PyTorch in fewer lines (2 comments): "The fact that it competes with PyTorch in so few lines speaks to the incredibly low incidental complexity imposed by Tinygrad." ([source](https://news.ycombinator.com/item?id=46424808)) "If your needs can be expressed as tensor operations or neural network stuff that tinygrad supports, might as well use that" ([source](https://news.ycombinator.com/item?id=47054871))

## What people warn about

- Dense, terse code hard to read (6 comments): "clearly formatting decisions were made to keep line count down over the flow of code" ([source](https://news.ycombinator.com/item?id=46443616)) "one look at pm_reduce_collapse and I’m glad I don’t have to work in this codebase." ([source](https://news.ycombinator.com/item?id=46441192)) "This operation still looks like a incomprehensible wall of code 40 years later" ([source](https://news.ycombinator.com/item?id=46442707))
- Little proven production use (2 comments): "I haven’t seen tinygrad used for any mainstream production project or thing of value, yet." ([source](https://news.ycombinator.com/item?id=47475335)) "I haven't heard many (any?) war-stories about a TinyBox in production" ([source](https://news.ycombinator.com/item?id=45987762))
- Big claims, little to show (2 comments): "Lots of words and weird analogies to say basically nothing." ([source](https://news.ycombinator.com/item?id=46423879)) "They're claiming a lot, but don't have much to show for it" ([source](https://news.ycombinator.com/item?id=49319509))

## Alternatives hackers recommend

- [PyTorch](https://hackerapproved.com/p/pytorch.md): 131 recommend
- [MLX](https://hackerapproved.com/p/mlx.md): 75 recommend
- [JAX](https://hackerapproved.com/p/jax.md): 49 recommend
- [OpenCV](https://hackerapproved.com/p/opencv.md): 46 recommend
- [TensorFlow](https://hackerapproved.com/p/tensorflow.md): 20 recommend

## Cite as

Hacker Approved, "tinygrad", https://hackerapproved.com/p/tinygrad/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).