Jane Street employees all have an AI twin that does their knitting for them
Investment banks are building digital employees. Jane Street is building 'AI Users.' The big difference is that these users uniquely correspond to each human employee at trading firm.
💥Follow us on WhatsApp for news alerts.💥
Speaking on the Signals and Threads podcast over the weekend, Jane Street engineer John Crepezzi said that the firm has built identities through which AI agents can act, "which are alternate versions of each of the people at the firm." Jane Street engineers code through Claude Code, Codex, or Jane Street's internal coding systems.
This was initially just a security classification so that Jane Street's systems would "apply the right set of rules" when an agent was acting on your behalf "because they know that the caller is an AI version of you." Since then, however, Crepezzi said that the firm has "built other systems that take use of this AI-concept."
One new use case is a code tracking system that monitors what code was written by these AI users and what was written by a human. Crepezzi didn't mention other use cases, but said the AI user structure is "show[ing] up in a bunch of other places too."
One of the ways that these users are called into action is through Jane Street's internal slack. Agents will start a project when prompted via chat and start writing the code for it without having to be prompted. Crepezzi said that "something like 45% of the total released features" start in this manner.
How is AI changing Jane Street
We've written about how AI is changing Jane Street's famous internships. CTO Ron Minsky has said in the past that there's increased focus on reviewing code rather than writing it due to the tendency of AI to write "uncanny" programs that look functional but are "super broken." Senior employees are coding less too, but they're not necessarily happy about it.
Speaking on the most recent podcast, Minsky likened coding to "knitting... you're not maybe solving any big ideas, but the craft of putting the code together" is an "artistic" pursuit. Crepezzi said that the focus is now more on "watching the code that it wrote and offering adjustments where you can."
There are benefits too. Crepezzi noted that AI users starting pull requests through Slack saves you from "a lot of toil involved in just creating a new pull request." Minsky said it also gives engineers time to work on the "code that you would’ve never bothered to write at all."
Crepezzi noted that systems design (the big picture stuff) is "still fairly protected." Minsky suggested that it has also become the bulk of work for Jane Street engineers; he was told by a "very strong engineer" that their work was previously 10% "thinking really hard and deeply about what to do," then 90% executing - now, Minsky said that engineer will offload the execution to an LLM and do "another one of those 10%" problems.
Have a confidential story, tip, or comment you’d like to share? Contact: WhatsApp: http://wa.me/442079977910 (+44 20 7997 7910), Telegram: @AlexMcMurray, Signal: @AlexMcMurrayEFC.88 Click here to fill in our anonymous form, or email editortips@efinancialcareers.com.
Bear with us if you leave a comment at the bottom of this article: comments are moderated intermittently by human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. You must take sole responsibility for comments you post on this site. We will take reasonable steps to weed out anything that we consider to be offensive or inappropriate.
Photo by Yuliia Pakhomova on Unsplash
