1. A language with no market is training its own speech models, the same week open models chased closed systems on video In New Zealand's far north, a Māori broadcaster is training speech models for te reo, a language too small for any vendor to serve, according to a report circulating on Hacker News.

2. Cars24 recovered 12% of dead leads by grading its AI on cost per successful task When a used-car shopper stops replying halfway through a chat, Cars24's system does not file the lead away as lost. A voice or chat agent picks the conversation back up.

3. The developers who trust LLMs least keep handing them more Software engineers who publicly concede the case against large language models are, in the same breath, granting those models wider autonomy.

In Brief

• Databricks reaches $188B valuation Databricks closed a funding round valuing the company at $188 billion, up from earlier marks, as investors rewarded its pivot from data warehousing to AI tooling. The company published research arguing open-weight models cut coding costs against closed rivals.

• Apple clears Apple Intelligence for China with Alibaba and Baidu Chinese regulators approved Apple Intelligence for launch using Alibaba's Qwen and Baidu models, a partnership rumored since last year. The deal lets Apple ship AI features in its second-largest market after months of delay.

• GPU lenders back inference chips in $400M loan Financiers who pioneered GPU-backed lending arranged a $400 million loan collateralized by inference chips. The shift signals capital moving from training hardware toward the chips that run deployed models.

• AI memory demand raises phone prices in India Surging AI demand for memory chips pushed component prices up, jolting India's smartphone market with higher prices and softer demand. Manufacturers are adjusting pricing and product strategy as the shortage reshapes the low end.

• San Francisco orders Apple and Google to pull nudify apps San Francisco's attorney general ordered Apple and Google to remove AI "nudify" apps from their stores, estimating the platforms earned millions in fees. The office says the companies profited from apps that generate non-consensual explicit images.

• Patreon blocks AI scraper bots with Cloudflare Patreon dropped its reliance on robots.txt requests and now actively blocks AI crawlers using Cloudflare. The company aims to stop bots from training models on creators' paid content without permission.

• TikTok tests AI likeness detection for creators TikTok began testing an opt-in tool that scans for unauthorized AI versions of a creator's likeness and lets them report matches. The test covers some US creators, mirroring a YouTube feature in development.

• Agility Robotics opens Digit training center in Fremont Agility Robotics opened a training facility for its Digit humanoid robots in Fremont, California, near Tesla's factory. The site will develop deployment for warehouse and logistics work.

• DoorDash opens command-line ordering for AI agents DoorDash launched a limited beta of dd-cli, letting developers and AI agents search stores, build carts, and place orders from the terminal. The tool targets software agents rather than human app users.

• Ex-DeepMind researcher raises at $300M pre-seed valuation Andrew Dai, a former DeepMind researcher whose work fed into ChatGPT, raised a pre-seed round valuing his startup at $300 million before shipping a product. He is betting on visual AI as the next frontier.

• Vertu sells a $6,880 AI agent phone to executives Vertu priced a luxury foldable with a built-in AI agent at $6,880, targeting executives. Testing found its AI workflows, battery life, and security fall short of the premium.

Read the full edition →