|
|
|
Aug 06, 2026
|
|
|
|
|
Supported by
|
|
|
|
|
|
|
Happy Thursday! Jeff Dean leaves Google in a major overhaul of the tech giant's AI leadership. Meta Platforms unveils new coding tools. DeepSeek resumes fundraising talks and plans to raise the prices of its AI models.
|
|
|
|
Google announced major changes to its artificial intelligence operations, with Demis Hassabis stepping aside as Google DeepMind’s chief executive to become its chairman and longtime AI scientist and executive Jeff Dean leaving the company to co-found a startup. CEO Sundar Pichai said Hassabis, who shared a Nobel Prize in 2024 for work on using AI for protein folding, will also become chief scientist of Google’s parent company, Alphabet. Koray Kavukcuoglu, who had been serving as Google DeepMind’s chief technology officer and Google’s Chief AI Architect, will lead Google DeepMind as senior vice president. The changes are likely to strike a blow to morale at Google DeepMind, which has already seen a string of departures in recent weeks, and could complicate Google’s efforts to catch up in the race to develop the best AI models, particularly if more researchers leave as a result. Google has fallen behind in the AI model race, particularly in coding. Its next flagship model has been delayed as rivals Anthropic and OpenAI have released new models that are strong on coding and cybersecurity. Google’s stock price dropped 4% on Wednesday’s announcement. Earlier this summer, star AI researcher Noam Shazeer left Google to go to OpenAI. John Jumper, who received the Nobel Prize alongside Hassabis for their work on the protein-folding project, called Alphafold, recently left the company to join Anthropic. The leadership change in DeepMind may not change as much in practice for Google’s Gemini development. Kavukcuoglu had already been running Google’s generative AI division, and was making more of the day-to-day calls on Gemini model development than Hassabis, researchers said. (Read our 2025 profile of Kavukcuoglu here.) Hassabis had been spending a lot of his time on external work, as Pichai noted in his announcement. Hassabis will also continue to lead Isomorphic Labs, DeepMind’s drug discovery spinoff. Dean, who has most recently been chief scientist for Google DeepMind and Google Research, joined Google in 1999 and is one of its most prominent employees. He is founding his new company with Sanjay Ghemawat, a senior fellow at Google who has been his longtime collaborator; Oriol Vinyals, a longtime Google DeepMind researcher who had been co-tech lead on Gemini alongside Dean and Shazeer; and Quoc Le, a fellow at Google who was a founding member of Google’s AI unit Brain alongside Dean. Their startup, Discovery Loop, will seek to automate AI research, a goal known as “recursive self-improvement” that has obsessed AI researchers who see it as the path to achieving artificial general intelligence, or systems that can perform any task as well as a human. The startup has received funding from Radical Ventures and Khosla Ventures, although the amount of funding and valuation wasn’t disclosed.
|
|
|
|
Meta Platforms unveiled a new coding model, Muse Code beta, on Wednesday alongside its latest AI model, Muse Spark 1.2. In a Threads post, CEO Mark Zuckerberg said Muse Code can delegate complex tasks to “sub-agents,” allowing developers to split programming tasks among multiple AI agents, instead of relying on a single AI agent. The release marks Meta’s latest effort to narrow the gap with Anthropic and OpenAI in AI coding tools, an increasingly competitive market where Anthropic’s Claude Code and OpenAI’s Codex have emerged as leading products. Alongside the announcement, Zuckerberg shared results showing Muse Code ranked second on Terminal-Bench 2.1, a widely used benchmark that measures AI agents’ ability to complete real-world software engineering tasks. It placed behind Anthropic’s Claude Code Opus 5 and ahead of OpenAI’s Codex GPT-5.6 Terra (max). On DeepSWE, a benchmark that evaluates AI models on software engineering problems, Muse Spark 1.2 ranked third, behind Anthropic’s Claude Code and OpenAI’s Codex. To improve its coding models, Meta has been pushing thousands of engineers to use its in-house coding agent, MetaCode, in their daily work and submit fixes when the system makes mistakes. An executive recently told engineers in a memo that feedback has already improved an earlier version of Muse Spark and boosted MetaCode’s performance on software engineering benchmarks.
|
|
|
|
Chinese AI developer DeepSeek has resumed its second funding round, which was paused for more than a week following a leak of the transcript of a confidential call between CEO Liang Wenfeng and investors, according to a person with direct knowledge of the talks. Earlier this week, DeepSeek shared a new term sheet for the funding round with investors to restart the discussions, according to the person. The company, which raised 50 billion yuan ($7.4 billion) in its first-ever funding round that closed in June, is looking to raise another 50 billion yuan in this new round at a higher valuation of 500 billion yuan ($74 billion), The Information previously reported. Chinese media outlet Caijing first reported the resumption of the funding talks. Separately, DeepSeek also said on Thursday that it plans a “significant increase” in the near future in the overall pricing for the application programming interfaces of its AI models, according to its website. The price hike is a major shift for DeepSeek, which has been offering competitive models at the most affordable prices. Liang previously told employees and investors that making big profits wouldn’t be a priority for DeekSeek, as the company’s mission is to make advanced AI accessible to everyone. Despite its rock-bottom prices, DeepSeek’s annualized revenue reached between $400 million and $500 million recently thanks to growing demand, The Information reported last month. DeepSeek didn’t explain why it’s raising the API pricing. The hike comes as the company is facing overwhelming demand for its recently launched V4-Flash, a small version of its flagship V4 open-source model. The model has gained massive popularity among developers who are looking for cost-efficient models that can power autonomous agents without breaking the bank.
|
|
|
|
Shopify stock jumped 17% on Wednesday after the ecommerce software service reported better than projected 34% revenue growth for the second quarter. While the growth rate was in line with Shopify’s first quarter, the company had projected revenue growth would slow to the “high twenties” in percentage terms. At the same time, Shopify’s operating income rose 68% to $488 million. Executives said the company was seeing “broad growth across our merchandises, geographies and sales channels,” adding that AI-powered search was helping boost shopping. Shopify president Harley Finkelstein said AI search was “particularly helpful to some of the smaller brands,” which make up most of Shopify’s merchant base. That’s a result of AI agents searching Shopify’s catalog to more precisely match what a shopper asks for than typically happens with traditional search.
|
|
|
|
| |
