Elad Hazan Recalls Two Decades of Online Convex Optimization — Born as a Job Hedge
HazanPrinceton · x · 2026-09-16
Princeton professor Elad Hazan publishes a personal recollection of two decades of online convex optimization (OCO):
- It began as a career hedge: as a grad student in computational complexity, surrounded by prodigies like Scott Aaronson and Irit Dinur, he feared losing the theory-job race and organized a finance study group as a backup — which is how OCO started.
- Key milestones: Zinkevich's OCO paper, Cover's universal portfolios, and collaborations with Satyen Kale and others shaped the field.
- On AI: he notes AI is now best at exactly what STOC/FOCS rewards, yet less good at inventing new frameworks — his personal read on AI's limits.
- The intimidating peers of his student years are all close friends today.
A mix of academic history and personal narrative, valuable to researchers and anyone interested in optimization theory.
Related event: Princeton's Hazan Reflects on Two Decades of Online Convex Optimization(3 posts)→
More from AGI Musings
- Hassabis and Legg launch DeepMind Institute for interdisciplinary AGI research — dhadfieldmenell · 2026-09-17
- Dean Ball flags anti-safetyists' mood-affiliation arguments on AI risk — deanwball · 2026-09-17
- One person can build an 'info nuke' in their basement, argues AI safety skeptic — Justin_Halford_ · 2026-09-17
- Writer loses $2m book deal after his African dialect was mistaken for AI text — theindependentonline · 2026-09-17
- Gary Marcus backs Stoller's critique: METR's HF incident probe ignored the humans, only watched the agents — GaryMarcus · 2026-09-16
- Altana CPO Kit Krugman: after automation, leaders' job is the messy work of unpredictable humans — every · 2026-09-16