1. OpenAI's IPO Prep Has a Body Count, and ChatGPT's Interface Is First OpenAI told the SEC it wants to go public. Inside the company, the consequences are already visible in the product.

2. Apple Built Its Privacy Brand on Doing AI Itself. The New Siri Runs on Google's Models. For years Apple sold a single idea about artificial intelligence: it would do the hard parts itself, on the device, where no one else could see your data.

3. A designer who used to throw out every AI mockup now builds prototypes instead of Figma files The designer had a rule about large language models: every time he reached for one, it let him down. Last year he tried Copilot and Cursor to tweak a game he had built.

In Brief

• NVIDIA and LG build an AI factory for robotics and autonomous driving NVIDIA and LG Group are constructing a dedicated AI factory to supply LG with computing infrastructure for training, simulation and deployment. The buildout targets robotics, autonomous driving, data center technology and GPU cloud services across LG's businesses.

• The UK funds a state-backed AI supercomputer to cut reliance on US chips The British government is financing a billion-dollar supercomputer to support homegrown semiconductor startups. The initiative aims to reduce dependence on US-designed silicon for AI workloads.

• DeepSeek claims V4 Pro outperforms GPT-5.5 Pro on precision tasks DeepSeek says its V4 Pro model beats OpenAI's GPT-5.5 Pro on precision benchmarks. The claim positions the Chinese lab against the latest US frontier release.

• Google adds Gemini 3.5 and Antigravity to NotebookLM Google upgraded NotebookLM with its Gemini 3.5 model and Antigravity features. The new capabilities are restricted to AI Ultra and enterprise accounts for now.

• Meta strips face-recognition code from its smart glasses app after WIRED report Meta removed face-recognition code that WIRED identified in the Meta AI companion app for its smart glasses. The company declined to explain the removal or say whether the feature returns.

• Malicious Microsoft packages hit developers a second time in weeks Attackers planted 73 packages carrying a self-replicating credential stealer that runs the moment an AI agent opens them. It marks the second such campaign targeting Microsoft packages in weeks.

• Apple lets Safari users generate browser extensions with AI Apple is addressing Safari's thin extension library by inviting users to build their own extensions through AI prompts. Safari has long trailed rivals because of Apple's strict development requirements.

• Apple adds prompt-based workflow building to Shortcuts Apple updated its Shortcuts app to let users describe a workflow in a prompt and have AI assemble it. The change removes manual step-by-step construction for automations.

• Microsoft AI chief Suleyman says superintelligence is near but won't take jobs Mustafa Suleyman argued that systems approaching superintelligence will not displace workers. He discussed his training approach and Microsoft's relationship with OpenAI in the interview.

• Ed Zitron argues AI progress is slowing Zitron published a detailed case that model improvements and the broader AI buildout are decelerating. The piece challenges the finances behind NVIDIA, Anthropic and OpenAI.

• GENEB benchmark exposes why genomic AI models resist comparison Researchers introduced GENEB, which tests frozen representations from 40 genomic foundation models across 100 tasks under one protocol. The work targets fragmented benchmarks that make cross-model claims unverifiable.

Read the full edition →