
Sign up to save your podcasts
Or


Jason Liu (Website, X, Github, Newsletter) is a technologist, consultant, teacher, and friend.
He spent the first part of his career as a machine learning engineer, mostly at Stitchfix, only to run into a wall: a hand injury that prevented him from being able to write any software for over a year. Fortunately, he's not so one-dimensional, and spent time reclaiming somatic experience in learning to free-dive, train Jiu-Jitsu, and return to the pottery practice he developed in art school, all while reckoning with big questions of ambition, purpose, and self-fulfillment. Since then, he's built a consulting practice helping modern AI companies better implement RAG (retrieval-augmented generation), avoid system design mistakes, hire elite talent, and build for an LLM-centric world. He maintains a large structured output library called Instructor with about 1m downloads per month, writes prolifically (which he does entirely via voice input with LLM editing, as we discuss), tweets semi-manically (he's grown to 30K followers on X with the simplest strategy I've ever heard anyone articulate—tweeting 30K times), and teaches courses on RAG and online consulting. Finally, my man can yap. He was a perfect first guest because he has no shortage of ideas but comes at nearly everything with a beginner's mindset.
Timestamps
References
 By Jackson Dahl
By Jackson Dahl5
2121 ratings
Jason Liu (Website, X, Github, Newsletter) is a technologist, consultant, teacher, and friend.
He spent the first part of his career as a machine learning engineer, mostly at Stitchfix, only to run into a wall: a hand injury that prevented him from being able to write any software for over a year. Fortunately, he's not so one-dimensional, and spent time reclaiming somatic experience in learning to free-dive, train Jiu-Jitsu, and return to the pottery practice he developed in art school, all while reckoning with big questions of ambition, purpose, and self-fulfillment. Since then, he's built a consulting practice helping modern AI companies better implement RAG (retrieval-augmented generation), avoid system design mistakes, hire elite talent, and build for an LLM-centric world. He maintains a large structured output library called Instructor with about 1m downloads per month, writes prolifically (which he does entirely via voice input with LLM editing, as we discuss), tweets semi-manically (he's grown to 30K followers on X with the simplest strategy I've ever heard anyone articulate—tweeting 30K times), and teaches courses on RAG and online consulting. Finally, my man can yap. He was a perfect first guest because he has no shortage of ideas but comes at nearly everything with a beginner's mindset.
Timestamps
References

531 Listeners

2,423 Listeners

1,081 Listeners

488 Listeners

161 Listeners

188 Listeners

133 Listeners

96 Listeners

102 Listeners

77 Listeners

502 Listeners

41 Listeners

133 Listeners

40 Listeners

42 Listeners