Interview
3Blue1Brown Talks Machine Learning with Jane Street
- Machine learning is projected to remain the primary driver of systematic trading, with the firm positioned at the beginning of an expanding curve where technical frameworks are expected to shift every five years, necessitating ongoing adaptation.
- Hiring strategies will continue to prioritize generalists with strong learning and adaptability skills, treating specific machine learning proficiency as a plus rather than a strict requirement, while interview processes will rely on novel math puzzles to assess problem-solving rather than vocational knowledge.
- The organization expects to justify a one-year investment in retooling current employees on new technologies, supported by a high retention rate where the median time until an employee leaves remains long enough to offset retraining costs.
- Structural evolution will follow an organic model where systems and hierarchies change to fit needs as they arise, utilizing a real-time auction for compute resources that fosters collective responsibility and allows anyone to bid without bureaucratic guardrails.
- A "one company" culture will persist, characterized by low-friction, cross-group communication without manager intermediaries for small-scope tasks, alongside a "low ego" environment where colleagues help each other and prioritize collective success over individual point-proving.
- The firm anticipates that the distinction between "traders" and "researchers" will remain blurred, with roles evolving to blend trading systems, machine learning infrastructure, and theoretical application, leading to confusion for outsiders accustomed to traditional financial titles.
- Future technical work will increasingly bridge abstract mathematical concepts like M estimators with concrete modeling problems, moving beyond off-the-shelf models to address the frontier of knowledge in the research community.
- Cultural norms emphasizing intellectual curiosity, such as ad hoc talks on physics and math, and flexibility for career breaks like pursuing a PhD mid-career, are expected to continue, though such breaks remain rare.
- The company expects to handle petabytes of market data and maintain complex internal tools for code reviews and releases, relying on a "global read-write" compute cluster to enable rapid troubleshooting and collaboration without formal reporting chains.
- Misconceptions from external trading firms regarding the firm's scale, systemic nature, and culture will persist, with the firm continuing to use analogies like "grocery stores" or "AI labs" to explain market making, despite their limitations in capturing the complexity of AI and prediction.
- The interview process is expected to remain characterized by "Type Two fun," featuring high stress in the moment but satisfaction in hindsight, while the "everyday re-evaluation" of tenure will continue to be driven by learning, agency, and enjoyment rather than fixed multi-year commitments.
- Specific historical technical elements like OCaml will remain a minor thread in the company's history, while the "original people" are expected to stick around to contribute to long-term cultural stability amidst changing technical landscapes.