Gergely Orosz · 8 min read · Product Strategy
Distributed databases with Peter Mattis
Read the original on newsletter.pragmaticengineer.com →
This episode features Peter Mattis, co-founder and CTO of Cockroach Labs and original creator of GIMP, discussing his journey from open source to Google to building distributed databases. He shares insights on keeping systems fast and reliable at scale, from Gmail's early storage architecture to distributed database tradeoffs. Peter also reflects on how AI has reignited his passion for coding and its potential to multiply the impact of domain experts.
Key Takeaways
- Ship despite competition: Peter nearly abandoned GIMP after hearing about a more ambitious competitor, but shipped anyway—the competitor never materialized. Founders should assume others have the same idea but recognize most won't execute.
- Early Gmail used B-trees for its storage layer, specifically for tracking email threads and unread counts, illustrating how foundational data structures underpin large-scale consumer products.
- AI has brought Peter back to hands-on coding after his role shifted toward management, and he believes AI can improve quality while multiplying the impact of domain experts rather than replacing them.
- Google's Colossus file system reduced storage overhead by 33%, demonstrating the massive efficiency gains possible through rethinking storage architecture from first principles.
- Perseverance through uncertainty is a recurring theme—from rejecting Google's first offer due to commute concerns to nearly shelving GIMP—highlighting how timing and persistence shape outcomes in tech careers.
Excerpt
Stream the latest episode Listen and watch now on YouTube , Apple , and Spotify . See the episode transcript at the top of this page, and timestamps for the episode at the bottom. Brought to You by • turbopuffer – The turbopuffer engineering team is doing something really cool: they are completely redesigning their storage architecture from first principles to make search faster, cheaper, and more reliable at scale. And they’re documenting all of it! Follow along their rewrite at turbopuffer.com/v3 • Linear – most of us work with agents in a “single-player” setting. Linear’s take is agent work…