OpenAI Scholars

OpenAI wants more people building AI who don't look like the current default: overwhelmingly male, overwhelmingly from a handful of elite computer science programs. So it's rolling out a scholars program, funding somewhere between six and ten people from underrepresented groups to spend three months doing nothing but deep learning, full stop.

The structure is simple on paper. Take a stipend, skip the day job, and dedicate three months to learning the field with actual mentors from OpenAI in your corner. No mention yet of exact dollar figures or which backgrounds count as underrepresented, but the intent is clear: lower the barrier to entry for people who don't have the luxury of unpaid time to teach themselves PyTorch and read papers.

The catch, and it's a good one, is that scholars have to ship something. Every participant open-sources a project by the end of the three months. That's a meaningful requirement. It turns a purely educational stipend into something closer to an apprenticeship with a public deliverable, which means the output isn't just a resume line — it's code other people can inspect, fork, and build on.

This is a small program by headcount. Ten people is a rounding error next to the thousands of engineers OpenAI itself employs, let alone the size of the field. But small cohorts with real mentorship and a public output requirement have a way of punching above their weight, especially when the alternative for a lot of talented people is simply never getting a foot in the door.