The Pulse #107: What does Amazon’s 5-day RTO mean for tech?
Amazon is the first Big Tech to mandate a strict 5-day return to office. What are the real reasons, will Amazon see a ‘brain drain’ as a result, and could other Big Tech companies follow?
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. OpenAI’s new model, LinkledIn opts everyone into GenAI training (save for those in the EU), Unity reverses runtime fees, what a startup “fire sale” looks like for employees, and more.
What does Amazon’s 5-day RTO mean for tech? The online retailer is the first Big Tech giant to mandate a strict, 5 days in the office policy. It says it wants things back exactly like they were before the pandemic. One reason is likely something less discussed: tax incentives from cities granted for staff being onsite, daily
Tech debt: further learnings. Two weeks ago, Lou Franco shared hard-earned lessons on keeping tech debt at bay. Lou shares two additional learnings: using tech debt payments to get into the flow, and how big rewrites need heavyweight support.
1. Industry pulse
OpenAI’s new model: o1
Last week, OpenAI revealed its latest model called OpenAI o1. Until now, LLM models got bigger in how much data they were trained on, but o1 is a change of direction. It spends more time “thinking” before it responds: basically, it chains several LLM operations one after the other. As software engineer Simon Willison puts it:
“One way to think about these new models is as a specialized extension of the chain of thought prompting pattern — the “think step by step” trick that we’ve been exploring as a a community for a couple of years now, first introduced in the paper Large Language Models are Zero-Shot Reasoners in May 2022.”
The new model is far more capable in its responses – even though these responses take more time to process – and brings a new dimension in how to use language models. Because the model spends more compute time on answers, it has become more expensive to use: o1 costs 100x as much as GPT-4o mini, and 3-6x as much as GPT-4.
This approach feels like the biggest recent development in LLMs this year. It’s worth withholding judgment until we use it more, and learn more about its capabilities and tradeoffs. AI keeps moving fast, and OpenAI is at the forefront.