OpenAI, Karpathy back Distill to fix ML's communication crisis

Machine learning has a communication problem. Papers are dense, results are opaque, and even experts struggle to understand what’s actually happening inside the models they build. A new publication launching today called Distill wants to change that — and it’s got some serious backing. Andrej Karpathy, formerly of OpenAI and now at Tesla, is on the steering committee, and Greg Brockman, OpenAI’s president, is helping fund a dedicated prize for clarity in the field.

Distill isn’t just another preprint server. It’s a publishing platform built specifically for explaining machine learning concepts using modern web technologies — interactive diagrams, embedded visualizations, the works. The idea is to make the underlying mechanics of algorithms actually legible, rather than buried in equation-laden PDFs that few people outside a narrow subfield will ever fully grok. Early examples already live on the platform: one piece digs into the finicky parameters of the t-SNE algorithm, another unpacks why generated images sometimes get those weird checkerboard artifacts, and a third literally looks under the hood of recurrent neural networks that produce handwriting.

The timing feels right. As models get larger and more complex, the gap between what researchers can build and what they can explain keeps widening. Distill is betting that better communication isn’t just a nice-to-have — it’s essential for the field to advance responsibly. The platform provides authors with tools that go far beyond what a static PDF or blog post can offer, letting readers manipulate variables and watch how networks behave in real time.

The Distill Prize for Clarity in Machine Learning, which Brockman is backing, adds another incentive layer. It’s open to outstanding explanatory work published anywhere, not just on Distill’s own platform. That’s a smart move — it signals they care about raising the bar across the entire field, not just building a walled garden. Whether the academic incentive structure actually shifts toward rewarding clear explanation over novel benchmarks remains an open question. But having names like Karpathy and Brockman attach their reputations and money to the effort certainly doesn’t hurt.