Why Tensors & Transistors exists
The gap between knowing how to code and understanding the whole system, and why the Spanish-speaking technical community needs to close it.
There is an uncomfortable asymmetry in technical education across Latin America: resources that teach you to use tools are everywhere, and resources that explain how they work underneath are scarce.
You can find a hundred tutorials on training a model with PyTorch. Far fewer on what a convolution kernel actually does in memory, why cache size decides throughput, or what that same model looks like once you quantize it to run on a microcontroller with 256 KB of RAM.
That second layer is what separates the person who executes from the person who designs. And it is almost always in English, behind a paywall.
What we publish
Three kinds of content, nothing else:
- Technical blog — long-form articles on architecture and design decisions.
- Tutorials — guides reproducible from start to finish.
- Paper summaries — research distilled to what you can use.
The bar
An article ships if it meets three conditions:
- It is verifiable. If there is code, it compiles. If there are numbers, the experiment can be reproduced.
- It explains the mechanism. Saying what to do is not enough; you have to say why it works.
- It cites its sources. Papers, datasheets and primary documentation get linked.
If it misses any of the three, it does not ship. We would rather publish less.
This site is open source. If you find an error, a pull request is welcome.