The Pulse #113: Engineering culture at Google vs Meta
Also: AI now generates 25% of code at Google; Deep cuts at profitable companies like Dropbox and Miro; Business booming at the likes of Google, Meta and Microsoft, and more.
The Pulse is a series covering insights, patterns, and trends within Big Tech and startups. Notice an interesting event or trend? Send me a message.
Today, we cover:
Industry pulse. AI creates 25% of new code at Google, GitHub Copilot ditches ChatGPT exclusivity, Arc browser struggles for mainstream attention, Microsoft engineer shrinks a huge Git repo by 94%, Mozilla could become an ads company, and more.
Engineering culture at Google vs Meta. Software engineer Roman Kirillov joined Google after 14 years and shares impressions on how the two companies differ, and how they are similar. A “monolithic” engineering culture at Meta, talking about moving fast — and then moving really fast and outages being more “everyday events” than at Google.
Deep job cuts at profitable companies: Dropbox, Miro. Both companies are growing and make a profit, but still did cuts around 20%. A reminder that laying off while being profitable is a real downside of going public or taking VC-funding.
Big Tech: business is booming; what recession? The largest tech companies are doing very well-busiess-wise, recent result show. They all seem to hire at a more conservative pace than how their revenue is growing, though.
1. Industry pulse
AI creates 25% of new code at Google
It has barely been two years since ChatGPT was released and took the world by storm. Google responded with Bard – now renamed Gemini – and today, the majority of software engineers use some kind of AI coding assistant.
Google has built and rolled out internal, custom GenAI tools, and uses these actively. This week, CEO Sundar Pichai shared during the company’s earnings call that “more than a quarter of all new code at Google is generated by AI, and then reviewed and accepted by engineers. This helps our engineers do more and move faster.”
Most focus is on the “25%” figure, but I’m more interested in “reviewed and accepted by engineers.” Google is known to use very heavy testing automation and modern shipping approaches like canarying, feature flags, experimentation, etc. Despite this, no GenAI code ships to production without human review! I think they’re right to not trust GenAI by itself.
Google is invested in making GenAI a success, and Gemini is a growing part of their business. So it’s safe to assume that Google is further ahead than its competitors in integrating Gemini (and GenAI) into its engineering pipeline. Google has been a trailblazer at innovating large-scale software engineering approaches, and GenAI code generation will be the norm across Big Tech. I suspect it already is.