Lauren Tan
Conference affiliation: SpaceX Super Intelligence
Lauren Tan, known as poteto, is a member of the React Compiler core team. She helped redesign how the compiler analyzes JavaScript so it can decide which calculations a React component can safely reuse on its next render. She also created pstack, a plugin that gives coding agents her methods for investigating a problem, making a change, and checking the result in a running application.
On this page
From web design to leading a studio engineering team
Growing up in Singapore, Lauren learned web design by inspecting page source, following tutorials, and trading ideas with designers online. As a teenager, she built websites for schools and small businesses. She went on to study finance, learned programming with Excel, and worked in interface design before becoming a software engineer.
At Netflix, Lauren led engineering for the Studio Programming team, which built software for scheduling and programming original content. When the manager who had helped her become a tech lead left, Lauren accepted a management role in UI Engineering, supporting the engineers who built tools for Netflix’s production studio. She made herself a promise to reconsider the job each year. She enjoyed choosing a team and coaching its members, but she missed the part of her day when she could sit down and make something herself. In February 2020, after four years at Netflix and two in management, she announced her return to hands-on engineering.
Lauren’s advice for prospective tech leads grew out of one-on-one conversations she offered to people from underrepresented groups. She asks engineers considering the role to think about whether they enjoy organizing work, communicating decisions, and helping colleagues deliver. She also maintains Hiring Without Whiteboards, a community-built directory where candidates can find employers that use pairing sessions, work samples, or discussions of real engineering problems in interviews.
What React Compiler can see
At Meta, Lauren led internal prototyping of React Server Components and added support for them to Relay’s compiler and runtime. By 2021 she had returned to management, this time supporting the React team.
In her own programming, Lauren had come to reconsider how much a design pattern could promise. Learning Elixir challenged habits she had picked up in Ruby and JavaScript. Functional techniques gave her more ways to solve a problem, while a plain loop could still be the clearest choice for her team. In This Time Is Different, she described giving up on the idea that choosing the right pattern could future-proof a codebase.
On React Compiler, Lauren worked with Joe Savona, Sathya Gunasekaran, and Mofei Zhang on the first rewrite after Xuan Huang’s prototype. They gave the compiler a control-flow graph, an explicit map of the paths execution can take through a function. That let it track values more precisely and infer their types. Consider a component that returns early when it has nothing to display and otherwise computes a theme. React requires hooks to run in a consistent order, so a developer cannot put a useMemo hook after that conditional return. The compiler can follow the path where the theme is needed and reuse the result when its inputs are unchanged. Its analysis also powers lint checks that flag bugs such as state updates that cause a render loop.
Lauren wrote the October 2024 beta announcement, explaining how teams on React 17 and 18 could adopt the compiler with a compatibility package and inviting developers into its working group. In October 2025, she, Joe, and Mofei announced React Compiler 1.0, the first stable release for React and React Native.
Turning recurring agent mistakes into code checks
Working on Grok Bot, Lauren paid attention to the examples a codebase sets for the next coding agent. In Cursor’s desktop application, she had seen agents add timers to work around a mismatch between VS Code’s imperative updates and React’s declarative rendering. Comments explained why each timer seemed necessary. As agents copied the pattern, the cost of creating and discarding timers made the interface lag.
Lauren’s practice is to turn a repeated correction into a check the agent can run. In Dune, the Electron client framework her Grok Bot team built, the directory structure tells agents which code belongs to the main process and which belongs to the renderer that draws the interface. A dependency check rejects imports across that boundary, catching main-process work that could otherwise stall the interface. The agent gets an error it can act on before a reviewer has to spot the mistake. Lauren explains the decision in All Code Is Slop (Except Mine).
Lauren added pstack to Cursor’s public plugin collection in May 2026. Its poteto-mode selects a workflow suited to the job, such as investigating a slowdown or implementing a feature. Its verification workflow produces project-specific instructions that an agent new to the application can follow. For a save operation, that means exercising the control and checking that the data persisted. Lauren also built a state-chart tool for pstack so a reviewer can examine which transitions a user should be able to make, then compare those with a recording of what the application actually did.
1 conference talk
Loading more talks and highlights…
