Welcome back!

Meta just took the gloves off. The company has launched Meta Superintelligence Labs and poached an entire star team of top OpenAI researchers, offering sky-high bonuses and unlimited compute. OpenAI is reeling, calling it a “break-in at our house”. Let’s unpack…

Today’s Summary:

• ⚡ Meta raids OpenAI for AGI talent

• 🖼️ FLUX open-sources image editing AI

• 🏥 Microsoft medical AI has 85% accuracy

• 🍎 Apple may outsource Siri’s brain

• 📊 New report maps model trends

• 🛒 Claude fails at vending business

• 🛠️ 2 new tools

TOP STORY

Meta escalates AGI race with talent coup

The Summary: Meta has launched Meta Superintelligence Labs, a new AGI-focused division, and poached at least 8 top researchers from OpenAI. Backed by high compensations and limitless compute, the new lab aims to leapfrog rivals in building superhuman AI. OpenAI’s leadership reacted with alarm, describing the losses as a “break-in in our house” while Sam Altman warned that Meta’s “mercenary“ tactics may cause long-term cultural damage in the industry.

Key details:

• Meta’s new superintelligence unit consolidates Llama, FAIR, and product AI under one roof with a mission to build superhuman-level intelligence

• Alexandr Wang (ex-Scale AI) and Nat Friedman (ex-GitHub) lead the new unit; Daniel Gross joins as well

• Meta is rumored to have offered eight-figure bonuses to top hires

• OpenAI has lost several top researchers to Meta : Trapit Bansal (key researcher on the o1 reasoning model), Shengjia Zhao (co-developer of ChatGPT and the core GPT-4 architecture), Jiahui Yu (co-lead of OpenAI’s multimodal team), Ji Lin (key engineer on GPT-4.1, GPT-4.5 and o4-mini), Shuchao Bi (GPT-4o voice mode lead), Hongyu Ren (contributed to GPT-4o and o3/o4-mini), Huiwen Chang (GPT-4o image gen)

• OpenAI’s internal Slack lit up with execs calling the exodus “a break-in at our house” while Altman slammed Meta’s approach as “mercenary and distasteful”

Why it matters: The AGI race just went from theory to trench warfare. Meta has turned compensations and compute access into offensive weapons, challenging the illusion that OpenAI’s mission alone can retain top talent. With infrastructure scaling and elite researchers on the move, the center of gravity in AI could shift. Whoever controls the top researchers, controls the next wave.

FROM OUR PARTNERS

Smarter Marketing AI Workflows

Unlock the Power of AI With the Complete Marketing Automation Playbook

Discover how to scale smarter with AI-driven workflows that actually work. This playbook delivers:

• A detailed audit framework for your current marketing workflows

• Step-by-step guidance for choosing the right AI-powered automations

• Pro tips for improving personalization without losing the human touch

Built to help you automate the busy work and focus on work that actually makes an impact.

Steal the Playbook

FLUX

FLUX.1 Kontext image editor is now open source

The Summary: Black Forest Labs has released FLUX.1 Kontext [dev], an open-source AI image editing model that challenges proprietary leaders like Gemini image editing. The 12B parameter model release gives developers a locally hosted option for state of the art image editing. It’s now freely available for research and non-commercial use (paid license for commercial use).

Key details:

• Lets you edit existing images using simple text prompts, making precise changes while keeping the rest of the image unchanged

• The model runs efficiently on consumer GPUs and supports both precise local and broad global edits

• Human evals show it beats Google’s Gemini Flash Image editing

• A new self-serve licensing portal now lets businesses license FLUX.1 tools for commercial use

Why it matters: Open models are catching up in generative image editing. FLUX.1 Kontext [dev] shows that open-weight models can match closed systems, avoiding locking developers into silos. NVIDIA’s backing hints at a quiet hardware-software alliance driving AI democratization.

Try it on HuggingFace

FROM OUR PARTNERS

A Smarter Way to Read the News

Fact-based news without bias awaits. Make 1440 your choice today.

Overwhelmed by biased news? Cut through the clutter and get straight facts with your daily 1440 digest. From politics to sports, join millions who start their day informed.

Sign up now!

MICROSOFT

Microsoft AI diagnoses complex medical cases with 85% accuracy

The Summary: Microsoft has introduced a diagnostic AI system that achieves 85.5% accuracy in complex medical cases. The new system, named MAI-DxO, simulates a team of expert clinicians and helps identify and assess complex illnesses more effectively.

Key details:

• AI agents in MAI-DxO work as a coordinated virtual panel to manage lab testing costs, reduce bias, and guide reasoning steps

• MAI-DxO hit 85.5% diagnostic accuracy on NEJM cases

• The Sequential Diagnosis Benchmark (SDBench) uses 304 cases to mimic real-time clinical workflows

Why it matters: MAI-DxO shows that AI can handle complex medical reasoning at scale and with precision. It avoids the trap of over-testing while reaching accurate diagnoses faster. That means better outcomes and a stronger benchmark for clinical decision support.

Research paper

QUICK NEWS

Quick news

• Apple may use Anthropic or OpenAI to power Siri

• The most important trends in foundation models (report)

• Anthropic’s Claude fails hilariously at running a vending business

TOOLS

🥇 New tools

• Songscription - Generate sheet music from audio clips or MIDI files

• Dyad - Free, local, open-source alternative to Lovable / v0 / Bolt