Welcome back to In the Loop, TIME’s new bi-weekly newsletter focusing on AI. If you are viewing this in your browser, consider subscribing to receive the next edition directly in your inbox.

Featured Individual: Ali Ghodsi, Databricks CEO

While OpenAI, SpaceX, and Anthropic are widely recognized as the three most valuable private companies in the U.S., Databricks, ranking fourth, maintains a somewhat lower profile. This enterprise, reportedly seeking new funding this week at a valuation of $134 billion, operates as a discreet yet essential force in the AI revolution. Databricks provides a platform enabling companies to consolidate disparate, complex data, manage it efficiently, and then use it to train AI models. Its CEO, Ali Ghodsi, is a charismatic computer science professor who continues to teach occasionally at U.C. Berkeley alongside managing his multi-billion-dollar firm. I interviewed him last week.

Regarding AGI — Artificial general intelligence has already been attained, according to Ghodsi. To support this provocative assertion, he reflects on his experiences from two decades ago in a computer science lab, where he and colleagues enthusiastically speculated on what AGI might entail: capabilities for conversation, reasoning, and identifying patterns within vast datasets. “Now that we possess it, it’s rather like, never meet your heroes,” Ghodsi remarked with a laugh. “It’s simply, okay, that’s not as impressive as we thought.”

However — Ghodsi’s perspective is strategically motivated. Although OpenAI is a major client, Databricks’ core business stems from the broader market of conventional companies not aiming for AGI, but rather seeking to leverage existing AI for data-related tasks. Ghodsi frames Databricks as capitalizing on the immense scope of this market. “If all AI advancement stopped today, I believe we have sufficient tools to continue our work,” Ghodsi stated. “We’re not striving to create a super-God… We are much more focused on: how can we make AI practical for organizations right now?”

On AI agents — “Agents” emerged as a prevalent but ultimately under-delivered buzzword of 2025. While AI experienced significant improvements this year, including enhanced capabilities for web searches and code execution, large AI models still encounter reliability issues that restrict their autonomous operation durations, meaning fully functional AI coworkers are not yet a reality. Ghodsi revealed that Databricks is developing a solution: assisting companies in training smaller, highly specialized agents using their proprietary data. These models prove both more economical to operate and more dependable for specific functions than leading AI models such as ChatGPT or Claude. “We can provide high quality for a specific task because we are taking a simpler approach, concentrating solely on the particular tasks you require assistance with,” Ghodsi explained.

If you have a moment, please complete our brief survey to help us better understand your interests and which AI topics you find most engaging.

Key Information: The Advent of AI Private Equity

For decades, major private equity firms have successfully employed a consistent strategy: acquire a struggling company, reduce expenses through layoffs or business process modernization, and then sell it for profit, before moving on to the next acquisition.

It was only a matter of time before this strategy was adapted for the AI era. OpenAI announced on Monday that it has acquired a stake in Thrive Holdings, a firm established earlier this year by Joshua Kushner’s Thrive Capital, which plans to purchase and revitalize struggling companies by boosting their productivity with AI.

Traditional private equity has historically benefited C-suite executives more than the average employee. Many workers at PE-acquired companies often complain that new management makes decisions that may temporarily inflate balance sheet figures but lack long-term strategic sense, or that they pay themselves substantial bonuses while reducing staff to a bare minimum.

Now, it appears OpenAI will be directly involved in a company perpetuating this established practice.

AI in Action

DeepSeek introduced a new iteration of its V3 model on Monday, which, based on certain benchmarks, is comparable to GPT-5 and Gemini 3.0 Pro. Anticipate increased concern in Washington and Silicon Valley regarding the extent to which Chinese AI is catching up to its American counterparts.

Our Current Reads

“,” by Robert Booth in The Guardian

OpenAI’s legal team demonstrated strong form here:

According to court documents filed at the superior court of the state of California on Tuesday, OpenAI asserted that “to the extent that any ‘cause’ can be attributed to this tragic event” Raine’s “injuries and harm were caused or contributed to, directly and proximately, in whole or in part, by [his] misuse, unauthorised use, unintended use, unforeseeable use, and/or improper use of ChatGPT.”