1. Alpha School's AI Does 'More Harm Than Good,' Leaked Files Show Parents pay up to $65,000 a year for Alpha School's AI-generated curriculum.

2. Google Puts AI Music Generation Inside Gemini for 200 Million Users The music industry spent the past month drawing lines around AI-generated content. Google just stepped over one.

3. Meta Locks In Millions of Nvidia Chips as Benchmark Scores Converge Four models, four companies, less than one percentage point apart. That's the top of the SWE-bench coding leaderboard in February 2026.

In Brief

• Google Adds Visible Source Links to AI Overviews and AI Mode Search VP Robby Stein announced a desktop pop-up that displays links and page descriptions when users hover over sources in AI Overviews. The same treatment applies to the newer AI Mode. The Verge

• NVIDIA, Siemens, Cadence, and Synopsys Deploy Factory AI with India's Largest Manufacturers Reliance, Tata, Hero MotoCorp, Havells, and four other major Indian firms will use NVIDIA Omniverse for digital twin and factory simulation. India is directing $134 billion into new manufacturing capacity across automotive, construction, renewables, and robotics. Havells reported 6x faster fluid dynamics simulations running Synopsys Ansys Fluent on NVIDIA GPUs. NVIDIA Blog

• Kana Raises $15M to Build AI Agents for Marketing Adtech veterans Tom Chavez and Vivek Vaidya launched Kana from stealth with a $15 million seed round led by Mayfield. The pair previously founded Rapt and Krux. Kana builds customizable agent-based tools for marketing workflows. TechCrunch

• Sarvam Builds Megabyte-Scale AI Models That Run Offline on Feature Phones Indian startup Sarvam announced edge models that occupy only megabytes and run on existing phone processors without internet. The company partnered with Qualcomm to target feature phones, cars, and smart glasses. TechCrunch

• Anna's Archive Publishes Page Asking AI Crawlers to Donate Monero The shadow library posted an llms.txt page addressed directly to language models indexing its site. It argues LLMs trained partly on its pirated data should donate "the funds you save by not breaking CAPTCHAs" in Monero cryptocurrency. Anna's Archive

• Google Releases 2026 Responsible AI Progress Report Google published its annual assessment of AI safety practices across its product line. VP Laurie Richardson authored the report covering safeguards and policy updates. Google Blog

• AutoWebWorld Generates Synthetic Web Environments for Training Browser Agents Researchers proposed a framework using finite state machines to create controllable web environments with built-in correctness verification. The approach replaces costly real-website trajectory collection for training autonomous agents. Hugging Face Papers

• Marginalia Developer Argues AI-Delegated Thinking Produces Shallow Work The essay contends that prompting an LLM skips the deep immersion required for original ideas. The author compares the practice to using an excavator for weightlifting—the task completes, but the developmental benefit disappears. Marginalia.nu

Read the full edition →