☕️ OpenAI prepares legal action against Apple

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

💥 OpenAI prepares legal action against Apple

✍️ Meta adds hand-gesture writing to Ray-Ban glasses

📉 Gmail cuts free storage to 5GB

💻 xAI unveils its first coding agent

📈 Ackman's Pershing Square buys Microsoft stake

Plus: 🎁 16 other news you might like, 🧰 6 tools, and 📚 4 papers.

💥 OpenAI prepares legal action against Apple LINK

OpenAI has hired an outside law firm to weigh legal action against Apple over their ChatGPT integration, with options including a formal breach-of-contract notice, though any move will likely wait until after its Elon Musk trial wraps up.

Announced at WWDC in June 2024, the deal placed ChatGPT inside Siri and the iPhone's Visual Intelligence feature, but OpenAI says the integration has been buried, its features are hard to find, and revenue trails projections.

Apple has its own complaints, including worries about OpenAI's privacy standards and annoyance at OpenAI's hardware push led by former Apple design chief Jony Ive and other ex-Apple executives now working with Sam Altman.

✍️ Meta adds hand-gesture writing to Ray-Ban glasses LINK

Meta is now letting everyone with its Ray-Ban Display smart glasses write messages using just hand gestures, with support across WhatsApp, Messenger, Instagram, and native Android and iOS messaging apps.

The hand-gesture writing works through the neural wristband included with the Meta Ray-Ban Display, a feature shown off at the glasses' announcement but held back from the launch lineup until a limited rollout this January.

Until now, the gesture writing was only in early access for WhatsApp and Messenger, so this update opens it up broadly and pairs it with other additions like display recording, expanded walking directions, and live captions.

📉 Gmail cuts free storage to 5GB LINK

Google is testing a change that drops the free Gmail storage offer from 15GB to 5GB for new accounts, with the missing 10GB only unlocked once users hand over their contact number.

The reduced quota, spotted by Android Authority, applies only to fresh signups, while existing accounts without a saved phone number still see the full 15GB shared across Drive, Photos, and Gmail.

Google did not publicize the shift on its support page until customers found it, later telling the outlet the experiment is framed as a security measure and is currently limited to a few regions.

💻 xAI unveils its first coding agent LINK

xAI has rolled out Grok Build, a terminal coding agent and CLI aimed at professional software engineering, putting the company in direct competition with tools like Anthropic's Claude Code.

The early beta is only open to SuperGrok Heavy subscribers paying $300 per month, who can install it from xAI's website and sign into their account to get access.

xAI describes the release as an early beta and says it will take user feedback to improve the product, which arrives after Elon Musk admitted the company had fallen behind rivals on coding.

📈 Ackman's Pershing Square buys Microsoft stake LINK

Bill Ackman's Pershing Square Capital Management has built a new stake in Microsoft, telling investors the market is undervaluing what the software company will gain from artificial intelligence.

The hedge fund began buying shares in February after Microsoft's fourth-quarter earnings disappointed, with cloud revenue growth missing forecasts and capital expenditures jumping, and Pershing will disclose the position in a regulatory filing Friday.

Ackman pointed to Azure and the M365 Office suite, which bundles the Copilot AI assistant at $30 a month, while calling fears about OpenAI's restructured deal and Microsoft's $190 billion 2026 spending plan overblown.

Other news you might like

• OpenAI brings Codex control to ChatGPT for iPhone and AndroidLINK

• What the jury will actually decide in the case of Elon Musk vs. Sam AltmanLINK

• Cerebras almost doubles in Nasdaq debut, topping $100 billion market cap after blockbuster IPOLINK

• Anthropic's Mythos AI outsmarted Apple's Mac security systemsLINK

• Spotify to adopt Apple’s new video podcast tech, offering creators easier cross-platform distributionLINK

• Microsoft pulls Claude Code licenses and pushes developers back toward its own AI toolLINK

• ChatGPT allegedly shared users’ query topics, user IDs, and email addresses with Google and Meta, new class action lawsuit claimsLINK

• Netflix wants to use generative AI to make animated shortsLINK

• Scientists create ‘sun battery’ using DNA-like molecules to store solar energy for yearsLINK

• Elon Musk’s SpaceXAI has been bleeding staff since its mergerLINK

• Not so locked in any moreLINK

• US firm launches ‘world’s first’ sensors to make machines see like humans with 1,640-foot rangeLINK

• A security researcher says Microsoft secretly built a backdoor into BitLocker, releases an exploit to prove itLINK

• Self-destructing $2,000 Nvidia chips will soon power tens of thousands of solar-run data centres hidden in lamppostsLINK

🧰 Trending tools

PHBench: uses emotional intelligence to analyze ProductHunt comment sentiment, helping you gauge how community reception actually impacts a product's ranking.LINK

Lensmor: a B2B prospecting tool that maps 160,000+ trade show exhibitors to verified decision-maker contacts, helping sales teams book meetings before events start.LINK

OpenHuman: a local-first, open-source AI agent that retains persistent memory across sessions while keeping your data off third-party servers.LINK

Gradient Bang: a browser-based AI game where you manage LLM-powered subagents via voice input, with a dynamically generated interface and custom programmable agents running in Vercel Sandboxes.LINK

Cline SDK: a modular framework for building agentic coding tools, extracted from the battle-tested Cline VSCode extension for extensible AI development workflows.LINK

Magic Notebook: a distraction-free writing app that keeps files local, supports DOCX, and removes setup friction for focused writing sessions.LINK

📚 Trending papers & reports

Large language models give less accurate, less truthful answers to users with lower English skills, less education, or non-US origins, making them least reliable for the people who need them most.LINK

LLM browser agents can be identified by the websites they visit just by watching their clicks and timing patterns, with classifiers reaching up to 96% F1 accuracy across 14 frontier models.LINK

AI co-mathematician scored 48% on FrontierMath Tier 4, a record among all AI systems tested, while helping researchers solve open problems and find overlooked research directions.LINK

AI language models can extract data from Spanish electricity bills with up to 97.61% accuracy, and crafting better prompts matters far more than tweaking technical settings, closing a 19 percentage point gap over basic approaches.LINK

Want to get the latest news differently? Find us on:

See you tomorrow for a new dose of ☕️ Techpresso!