Chris Li — End-to-end product builder

About Me

2025: Muse Spark & Muse ImageData & Acceleration TL Building human and synthetic datasets w/ researchers & annotators. Acceleration tooling.
2023: Meta AIMetaFounding engineer & TL Launching Meta AI from scratch and early product growth.
2019: MessengerFacebook Making Messenger fast and reliable through data sync.
2017: New product developmentebay Software engineer on two zero-to-one projects.
2015: Seller shipping time predictionebay Predicting how long items take to arrive. Feature engineering + hyperparameter experimentation.
2013: Apache Hadoopebayopen source contributor Contributed congestion control systems for large (at the time) HDFS clusters.
2012: Apple Building software in a hardware lab.
2011: Sold all my bitcoin
2011: Bought my first bitcoin

Growth is as much about how to unlearn what used to work as it is about learning new skills
- Now is a time to rethink what we think we know about software engineering
- Psychological blockers we have are things which were adaptive at one point
As long as coding models progress, the best teams will constantly adapt and rethink their core processes
- The most effective ICs review work like engineering directors
The best teams have a fun “core game loop” that leads to continual improvement
- Applies to health, fitness, and product development
Judge performance on outcomes but apply learnings to process
- Life is noisy: good decisions can lose and bad ones can win
- The process is consistent enough to improve
Pay attention to the left side of the technology adoption curve
- It takes years before a trend reaches mainstream acceptance
- What are we still doing today that is already known by others to be sub-optimal?
The most persistent disagreements are about incentives, not information
- If just information, sharing it ends the argument
- If incentives, the reasons regenerate as fast as you refute them
Subtraction is systematically underrated for large companies
- Removing is harder to take credit for than adding
- Local performance review optimization leads to emergent cultural biases