Welcome again to In the Loop, TIME’s fresh bi-weekly newsletter focusing on AI. For those reading in their web browser, consider subscribing to receive future editions directly in your email.

What to Know: My day with the robots

Last summer, I was in the peculiar situation of observing a humanoid robot attempt to do laundry. It crouched next to a washer-dryer, extended one hand into a laundry basket held by its other hand, and placed some garments into the drum. However, on two consecutive attempts, it dropped an item of clothing and was unable to retrieve it. An engineer used a litter-picker to sheepishly move the dropped fabric behind the appliance, out of my view. Nevertheless, the robot’s shortcomings were too evident to conceal.

Promising Venture — I was at the main office of Figure AI, an ambitious startup aiming to become the largest company globally. The previous week, they unveiled the Figure 03, their latest humanoid, to considerable excitement within the industry. This introduction underscored how the widespread optimism surrounding the AI surge is now influencing robotics. Many companies are currently vying to develop the world’s first functional humanoid robot, certain that recent breakthroughs in AI have transformed what was once a distant dream into a tangible possibility. Figure AI has emerged as one of Silicon Valley’s most talked-about—and financially significant—startups, having recently secured $1 billion, which values the company at $39 billion.

Featured on Cover — The Figure 03 was featured on TIME’s cover last week, included in our Best Inventions of 2025 edition. I authored the piece that accompanied this cover. I view the article’s aim as dual. Primarily, it intends to introduce much-required firsthand accounts into the ongoing discourse about Figure’s robots, which are not yet commercially available. Up to now, the public could only gauge their abilities through expertly filmed YouTube videos and corporate blog entries. My investigation suggests that Figure’s robots are not as dependable as these promotional materials might lead one to infer.

Future of Robotics — Secondly, my objective was for the article to address the likelihood that humanoid robots could still be a near-future reality. Subscribers to this newsletter are likely acquainted with AI scaling laws—the principle that an AI’s proficiency improves proportionally with the volume of relevant data and computational resources it processes. For instance, GPT-2 was quite underwhelming upon its 2019 launch, yet just four years later, GPT-4 astonished everyone. The implication is that AI can acquire new competencies with surprising speed. Figure asserts—with persuasive evidence—that its robots are following a similar developmental path. Therefore, considering the vast sums being invested in advancing robotics, we ought to seriously anticipate the potential for humanoids to join the labor force within years, rather than decades.

You can access my complete Figure 03 article . I encourage you to email and share your thoughts!

Who to Know: Hock Tan, Broadcom CEO

OpenAI extended its series of agreements with semiconductor manufacturers on Monday, announcing a collaboration with Broadcom to develop and implement AI accelerators totaling 10 gigawatts.

Broadcom possesses significant influence in the AI sector. It collaborates with major technology corporations to create bespoke, proprietary chips; for instance, it has an enduring partnership with Google for the creation of the search giant’s Tensor Processing Unit—an AI chip that Google employs as a substitute for Nvidia’s GPUs.

It appears OpenAI is now seeking a comparable arrangement. “By creating its own chips and systems, OpenAI is able to integrate insights gained from developing advanced models and offerings directly into the hardware, thereby enabling unprecedented levels of functionality and intellect,” OpenAI stated in a blog entry on Monday.

For Hock Tan, Broadcom’s Chief Executive Officer, the agreement with OpenAI represents merely the newest achievement in a prosperous year where his company’s shares had already surged over 30% even prior to the increase spurred by the OpenAI revelation. This positions Broadcom as the seventh most valuable enterprise globally.

AI in Action

On Monday, California Governor Gavin Newsom enacted several new AI-related bills, among them SB 243, legislation introducing provisions aimed at safeguarding children from “companion” chatbots. This law, conceived by state senators Steve Padilla and Josh Becker, is scheduled to become effective on January 1, 2026, and mandates that companies providing companion chatbots enforce age verification and inform the state about their procedures for handling suicide and self-harm risks.

“While these corporations possess the capacity to lead global innovation, it is our duty to guarantee that this does not compromise the well-being of our children,” Padilla declared in an address delivered shortly before the bill’s approval.

What We’re Reading

— authored by Noah Smith on Substack

A rising discussion questions if we are experiencing an AI bubble. Nevertheless, the majority of AI researchers indicate that AI functionalities are steadily improving, and corporations such as OpenAI, Microsoft, Google, Amazon, Meta, and Nvidia intend to keep allocating hundreds of billions towards data center expansions throughout the rest of the decade. This positive outlook might calm investors’ immediate apprehensions, even with ongoing worries about whether AI will generate sufficient investment returns to warrant the progressively debt-financed expenditure. However, economist Noah Smith contributes a stark insight: “Examining the history of industrial bubbles, and new technologies in general, reveals that for AI to trigger a collapse, it doesn’t need to completely fail. It merely needs to slightly underperform the expectations of its most fervent supporters.”