Welcome back to Week in Review! We’ve got tons of cool stuff for your reading pleasure this week: Jeff Bezos backs EV startup Slate; a Meta whistleblower accuses the company of collusion; Waymo may use interior camera data; and much more. Let’s get to it!
I want this: Slate, an EV startup, has the ambitious goal of building an affordable two-seater pickup truck for the attractive price of $25,000. It’s amassed a sizable war chest in service of that goal, backed by Jeff Bezos, and hopes to get its vehicle into production as soon as late 2026.
China collusion: Sarah Wynn-Williams, Facebook’s former head of global public policy who wrote a book about her time at Facebook, testified before the U.S. Senate this week. Her testimony was spicy, as you can imagine. According to Wynn-Williams, Facebook, now known as Meta, worked directly with the Chinese Communist Party to “undermine U.S. national security and betray American values,” she said.
Wait, what? Trevor Milton, the Nikola founder who was recently pardoned after being convicted of securities fraud, is trying to buy the assets of his former company out of bankruptcy. It’s unclear if any other parties submitted bids for Nikola’s assets.
This is TechCrunch’s Week in Review, where we recap the week’s biggest news. Want this delivered as a newsletter to your inbox every Saturday? Sign up here.
News
Say “cheese”: According to an unreleased version of Waymo’s privacy policy, the self-driving car company is planning to use data from its robotaxis, including video from interior cameras tied to rider identities, to train generative AI models. Users will apparently be able to opt out.
Back, back, back again: President Trump on Tuesday signed an executive order that endorses coal for data center power. The government will be directed to designate coal as a critical mineral and prevent the closure of some coal-fired power plants, requiring them to continue operating.
How to be rich: A breach of Berkshire Hathaway-owned private jet company NetJets has revealed some information about how flight attendants should serve Elon Musk on planes. According to the guide, Musk apparently isn’t “interested in conserving fuel” because he “wants to fly as quickly and as direct” as possible. He also likes to keep the cabin at a frigid 65 degrees.
Scooping up talent: OpenAI’s former CTO Mira Murati’s new AI venture, Thinking Machine Labs, has hired some prominent names in the field to be advisers — Bob McGrew, previously OpenAI’s chief research officer, and Alec Radford, a former OpenAI researcher behind many of the company’s more transformative innovations.
Dropping Dropbox: Dropbox chief customer officer Eric Cox, who joined the company in 2023, is stepping down, according to an SEC filing. It’s not clear yet who will replace him.
Got HVAC on my mind: Nest co-founder Matt Rogers knows about rolling with the punches. “Nest is not necessarily doing everything that I set them out to do years ago,” Rogers told Tim De Chant. “It’s one of the things when you sell a company.” But Rogers hasn’t been able to shake his obsession with HVAC.
Stick a fork in it: At a summit exploring how AI will affect education, U.S. Secretary of Education Linda McMahon referred to AI as “A1,” like the steak sauce. During a panel, she said “AI” at first, but became increasingly less consistent, leading us to believe that she knows the difference and it was just a slipup. A tasty, tasty slipup.
Analysis
$$$$$: AI itself is super expensive for companies to run, but we’re finding out that testing these models can be pretty costly, too. Evaluating OpenAI’s o1 reasoning model, for example, costs $2,767. Benchmarking Anthropic’s recent Claude 3.7 Sonnet “hybrid” reasoning model on the same set of tests is $1,485.35. Compare that to how much it costs to evaluate OpenAI’s o1-mini ($141.22) and Claude 3.7 Sonnet’s non-reasoning predecessor ($81.41). Kyle Wiggers looks at why benchmarking is getting more expensive as models get bigger and more complicated.
