Hiring at $850K a Year: The Job Is Watching a Group of AIs Collaborate

Lenny Rachitsky posted on X a list of currently highest-paying open roles, 25 positions, with data from LennysJobs. Salaries are base salary only and exclude equity, bonus, and signing bonus. At the top is Anthropic's Staff Research Engineer, Multi-Agent Scaling, at $500K-$850K. OpenAI's Research Scientist / Research Engineer, Search Integrity is second, at $380K-$850K.

First, the money in the list:

The full list and application links are in Lenny's original post and LennysJobs.

There is something easy to miss in this table: high pay is not only in model training. Product design, engineering management, brand, product marketing, legal engineering, payments risk, Kubernetes, and developer productivity are all in there. Ramp's Head of Brand ranks at $309K-$424K, and OpenAI's Product Marketing, Messaging Lead is $381K-$423K. Harvey's Head of AMER Legal Engineering is $400K-$450K. Companies pay according to scarcity, not according to whether the role title contains AI.

The Anthropic role capped at $850K has a very specific job description: design and run large-scale agent team experiments, study how performance, cost, and coordination change as the number of agents, compute budget, and task length grow together, identify bottlenecks, build platforms and evals, and debug failures that only appear at scale. Put plainly, one part of the job is figuring out why a group of agents starts dragging each other down once scale increases.

Lenny himself added in a reply: he limited the number of Anthropic and OpenAI roles, otherwise more than half the list would be taken up by those two companies. Commenter Greg Glazewski noted that this is only base salary, and the Anthropic role could be higher with equity. Commenter Jatin Batra asked more directly: who is this list for, since people earning this much probably don't have time to scroll X. That complaint isn't harsh. High-salary lists easily turn into screenshots, but roles being open does not mean they are easy to get, nor does it mean the money is in hand. Commenter Emon Datta added that public salary ranges are target ranges, not guaranteed liquid wealth.

For developers, what is more worth watching is how role requirements are changing. A position like Multi-Agent Scaling wants people who can design experiments, build systems, stay skeptical about numbers, and start from ambiguous problems. It does not require published papers, nor does it require multi-agent or reinforcement learning experience. For ordinary readers, remember one thing first: this is base salary for U.S. roles, in USD per year, excluding equity and bonus, and before living costs, visas, and team fit are factored in.

The truly useful part of this list is seeing which capabilities the market is pricing. Not all high-paying roles are training models; some are fixing infrastructure, some are managing growth, and some are turning brand and legal engineering into scalable systems. Money flows to scarcity, not to nouns.