Plus: Claude Cowork goes mobile, Meta pulls its deepfake feature after backlash, and SK Hynix makes Wall Street history
Sponsored by
Build apps without the coding headaches. Rork transforms your ideas into mobile apps through AI-powered conversations, helping founders move from idea to launch faster.
Create your app with RorkWelcome back to the world's #1 AI newsletter.
Today, we've got OpenAI unleashing GPT-5.6 on the world (with government approval), Apple going scorched earth with a trade secrets lawsuit against OpenAI, Meta launching and then killing an AI deepfake feature in the same week, and Anthropic making Claude Cowork available everywhere.
It's been one of those weeks where every major AI lab was doing something. Let's dig in.
On this day in 2025…
Google introduced "Deep Search" in AI Mode, a feature that could conduct hundreds of searches and compile comprehensive cited reports in minutes. The same week, video game actors secured a landmark agreement with studios, winning AI consent and disclosure protections alongside a 15% pay increase.
🚀 OpenAI Publicly Releases GPT-5.6 Family
OpenAI dropped its GPT-5.6 Sol, Terra, and Luna models after a two-week government-restricted preview. Sol is being called OpenAI's "best coding model yet," claiming to beat Anthropic's Fable 5 on the Coding Agent Index while using half the tokens and costing a third less. Alongside 5.6, OpenAI also launched GPT-Live — full-duplex voice models that can listen and speak simultaneously, making ChatGPT conversations feel genuinely human for the first time.
⚖️ Apple Sues OpenAI Over Alleged Trade Secret Theft
In a dramatic reversal of their 2024 partnership, Apple filed a federal lawsuit alleging OpenAI orchestrated a "pattern of theft" through former Apple employees. The suit names OpenAI's Chief Hardware Officer Tang Tan and engineer Chang Liu, accusing them of smuggling out prototypes, CAD files, and confidential specs. Apple claims OpenAI even coached departing employees on evading security procedures.
📱 Meta Kills Instagram AI Feature After Deepfake Backlash
Meta launched Muse Image on Tuesday and pulled its most controversial feature by Friday. The tool let users generate AI images using public Instagram accounts by simply @-mentioning them — no consent required. SAG-AFTRA called it "an utter miscalculation of public sentiment," and talent agencies pushed back hard. Meta admitted the feature "missed the mark."
💻 Anthropic Brings Claude Cowork to Mobile and Web
Claude Cowork, Anthropic's agentic assistant, is no longer tethered to your desktop. Now running in the cloud by default, Cowork lets you start a task on your laptop, check progress from your phone, and pick up finished work from any browser — even with your computer off. Anthropic revealed that 90% of Cowork sessions aren't even about coding. The coding agent wars are officially spilling into the rest of the office.
⚡ Mindwalk lets you replay coding-agent sessions on a 3D map of your codebase. Think of it as a visual debugger for AI-assisted development — you can trace exactly what your agent touched and how it navigated your project files. Trending on Hacker News with 146+ upvotes.
🧑💻 Iroh (Mesh LLM) brings distributed AI computing to the masses. Built on the iroh networking library, Mesh LLM lets you pool GPUs across machines to run large language models without a centralized server. It hit 332+ upvotes on Hacker News as developers explored true peer-to-peer AI inference.
Coming in Hot
AI email writer that sounds like you
AI Architecture & Interior Design Platform for Stunning Visualizations
Open-source AI compliance for SOC 2 and ISO 27001
Durable workflow orchestration for AI and backends
AI-powered brainstorming and mind mapping for teams
AI coworker that lives in Slack and does real work
AI-powered creative workspace for teams
Turn ideas into AI-powered apps from chat
All-in-one AI assistant right in your browser
AI-powered UGC video ads in minutes
Submit your AI tool to the #1 AI tools marketplace.
What it does: ForthWrite is an AI email writing assistant for Gmail and Outlook that learns your personal writing style from your sent emails. It generates context-aware email drafts in your voice, improves with every edit you make, and supports multiple AI models or your own API key (BYOK).
Why It's Worth Your Time
Pricing: Free plan available with 10 AI-generated drafts per week. Paid plans start at $15/month (Standard), $30/month (Pro), and $59/seat/month (Teams), with 14-day free trials and a 60-day money-back guarantee. Annual billing discounts are available.
What it does: ArchyBase is an AI-powered architecture and home design platform that transforms photos, sketches, and floor plans into photorealistic interior designs, exterior makeovers, landscape concepts, 2D floor plans, and 3D visualizations in seconds. It combines multiple AI design tools into a single workflow for homeowners, architects, interior designers, and real estate professionals.
Why It's Worth Your Time
Pricing: Free plan available with 4 AI design generations. Paid plans start at $15.92/month (billed annually) for 100 AI generations, with Plus at $31.92/month and Professional at $71.99/month, alongside one-time credit options.
🎙️ ChatGPT's Voice Mode Gets a Brain Upgrade
GPT-Live's full-duplex architecture means ChatGPT can now listen and talk at the same time — and it knows when to stay quiet. OpenAI thinks voice could become the primary interface for computing. Those rumored AI earbuds suddenly make a lot more sense.
💰 SK Hynix Raises $26.5B in Biggest Foreign US IPO Ever
The AI chip boom's biggest Wall Street moment: SK Hynix, Nvidia's key memory chip supplier, debuted on Nasdaq with a $26.5B offering that was 7x oversubscribed. The money's going straight to new fabs to meet insatiable AI demand.
