Mark Zuckerberg is making a big bet on getting smaller. It can be called “startup mode.”
Meta's new Superintelligence unit includes a secret group of hotshot adoption called the TBD Lab, shown by a recent internal memo.
The unit represents a part of the meta workforce of over 70,000 employees. Many of the members, including leader Alexandre Wang, were hired by buzzy AI startups.
“I've been a little more certain about the ability of a small, talented team to be the best structure to drive frontier research,” Zuckerberg said in Meta's latest revenue call.
“That's a little bit of a different setup,” Meta CEO said of the Superintelligence group, in contrast to the corps of engineers needed to manage powerful algorithms like Facebook's NewsFeed.
It's not just meta. The Gospels of a Small Team are rapidly spreading across Silicon Valley, where DNA-free startups are available.
From big technology to bare bone startups, a group of star workers pushes the frontier, aided by the leader's desire to get through AI and the bureaucracy.
With small teams entrenched following the hiring boom in the pandemic era, it remains unclear whether the model will work as big as possible for startups.
It has not stopped the major players from trying.
Cut red tape and ship faster
It's more than just a big skill that Micro spends a minute.
Kashish Gupta, co-founder of Haitoush, a marketing startup at San Francisco AI, which was valued at $1.2 billion earlier this year, strongly advocates for a small team approach.
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Hightouch is raising more than $132 million, while avoiding a major expansion of its head count with a total of 55 engineers, Gupta said.
Startups expect engineers to become self-starters who build products without the need for micromanagement from their superiors. For example, the launch of a major new AI agent has only four engineers working on it.
“Ultimately, the people who prioritize work are the engineers themselves or the people they work with,” Gupta said.
Gupta is not alone in his enthusiasm.
“A smaller team is better,” reads a manifesto written by investor and former Github CEO Nat Friedman on his personal website.
“Faster decisions, fewer meetings, more fun,” Friedman writes, claiming that many tech companies are overworking 2-10 times more.
Meta declined to comment on Business Insider about its management approach. Friedman, who joined Meta's super intelligence efforts in June and headed the division that integrates AI into the product, did not respond to requests for comment.
Zuckerberg has been tinkering with Meta's Org Chart and Head Count for many years, touting “Frath Is Better” and “Leaner faster” as part of Meta's 2023 efficiency.
His latest paper: in some cases, even small ones are better.
Trends overcharged by AI
AI breakthroughs often rely on small groups of researchers.
The basic architecture behind most modern, large-scale language models was featured in a 2017 paper, “Caution is required.” This is written by eight researchers, primarily from Google. Also, a much smaller pool of AI researchers is hired to begin with.
Both factors help explain why Meta reportedly tried to poach researchers with an astounding nine-digit compensation package.
“Based on my experience, advances in AI really rely on breakthroughs coming from just a few people,” AI engineer Yangshun Tay previously told Business Insider.
“It's not much needed. It's just that smart, cream-off cream people have a big breakthrough and a very disproportionate impact,” he said.
Small teams aren't necessarily magical fixes
The big question right now is whether these startup-style teams can push giant tech companies like Meta to the cutting edge.
Small teams with broad rights can face different challenges. One reason is that groups representing small dots on a vast ORG chart may still be seen by the legal, HR and recruitment teams of companies, says Elliott Parker, CEO of Alloy Partners, which will help large companies create startups.
Beyond that, special units within an organization can make legacy workers feel stuck on the side “not where the stars go.”
We previously reported that the arrival of a newly luxuriously compensated Super Intelligence Unit at META has caused tensions and threats of resignation from existing researchers.
There may also be challenges related to potential overlap that lead to multiple small teams doing the same thing. Google compares this issue with “slime-type,” but Meta has already reorganized its AI division multiple times. I've been melting two units in the last four months alone.
Meta's latest Reorg might do very well. However, Parker says innovation teams within large companies rarely change their parent organizations.
Small teams often offer useful new products, services and increased efficiency, but he said it's not the “basic transformation that companies often want.”
Zuckerberg has a sunny take. Last month's Meta revenue call, he sounded bullish and at least the right approach to maintaining his advantage in the AI race.
“For major research on super intelligence, you really want the smallest group that can hold the whole thing in your head.
