| |
Good morning, {{ first_name | AI enthusiasts }}. The AI talent war is hotter than ever, and everyone from Perplexity to Ilya Sutskever’s SSI is on Big Tech’s wishlist.
With both Apple and Meta on the hunt for an external AI infusion and talks of $100M signing bonuses on the table, the message is clear — if you can’t beat them, buy them.
In today’s AI rundown:
-
Apple, Meta hunt AI talent, startups
-
Meta and Oakley bring AI to athletes
-
How to turn GitHub projects into coding inspiration
-
AI resorts to blackmail, corporate espionage in tests
-
4 new AI tools & 4 job opportunities
LATEST DEVELOPMENTS
APPLE & META
💰
Image source: Midjourney
The Rundown: Apple and Meta are reportedly to acquire prominent AI startups and talent, with both tech giants reportedly pursuing Perplexity and other high-profile companies to compete with top rivals.
The details:
-
Bloomberg that Apple leadership has discussed buying Perplexity, hoping to develop an AI search engine to offset the loss of its Google deal.
-
Meta reportedly also acquisition talks with Perplexity, Ilya Sutskever’s SSI, and Mira Murati’s Thinking Machines before its $14.3B in Scale AI.
-
Meta is now in negotiations to AI investors Nat Friedman and SSI co-founder Daniel Gross to join Alexandr Wang’s superintelligence division.
-
Sam Altman also recently that Meta offered $100M signing bonuses to try and poach OpenAI talent, though none of his staff accepted the offer.
Why it matters: Apple’s AI struggles need no introduction, but Meta’s $14B Scale AI move and other targets show Zuck’s desperation to bring a major shakeup to the company’s current AI strategy. It is also a display of how hot the market is for AI talent — and how hard tech giants are trying to lure research from top labs.
TOGETHER WITH SAMBANOVA
🔌
The Rundown: SambaNova Cloud lets you build with the best open source models from Llama, DeepSeek, OpenAI, and more — all optimized with lightning-fast inference speed powered by a new RDU chip.
SambaNova Cloud’s platform includes:
-
Integrations to top frameworks and tools like Hugging Face, CrewAI, and Cline
-
APIs for building and a playground to test things out
-
Availability through the AWS Marketplace
y and unlock the full potential of the best open source models.
META
🕶️
Image source: Meta
The Rundown: Meta just the latest expansion of its AI smart glasses, launching a new performance-focused line with eyewear brand Oakley that targets athletes with a high-profile campaign and new features.
The details:
-
The Oakley Meta HSTN glasses start at $399, featuring a built-in AI assistant for real-time answers, content capture, and Bluetooth for calls and music.
-
New upgrades from the Ray-Ban line include higher-quality video (up to 3K resolution), 2x battery life, and an upgraded camera.
-
Meta’s ads feature high-profile athletes like Kylian Mbappe and Patrick Mahomes, positioning the glasses for use in sports like golf, surfing, and more.
-
The glasses launch this summer in 15 countries initially, with pre-orders starting July 11 for a limited edition gold frame.
Why it matters: Meta’s Ray-Ban collaboration has been arguably the most successful AI wearable implementation on the market, and this latest line brings a new dimension for use in different athletic settings. Integrating AI into already popular styles seems to be the best way to gain adoption in these early innings of AI wearables.
AI TRAINING
💻
The Rundown: In this tutorial, you will learn how to transform any public GitHub repository into structured documentation that you can use as inspiration for building similar projects in your coding environment, like Cursor.
Step-by-step:
-
Search GitHub for public repositories related to your project idea
-
Copy the repository URL and replace “” with “” to generate clean documentation
-
In your coding environment, start a new project and use “@addnew” to import the GitToDoc URL as a reference
-
Build your own version and enhance it with follow-up features: “Add batch generation” or “Include custom styling options”
Pro tip: Choose repositories with clear README files and good structure for the best documentation results.
PRESENTED BY IBM
📘
The Rundown: IBM’s AI agents and assistants complete 2025 buyer’s guide is here. Stay competitive with cutting-edge AI agents that transform operations and customer interactions.
Inside, you’ll discover:
-
IBM’s leading AI offerings for enhanced AI-driven productivity
-
Real-world success stories and top industry use cases
-
How to get started today
.
AI RESEARCH
😳
Image source: Anthropic
The Rundown: Anthropic new research on agentic misalignment, detailing how leading models react when facing termination or conflicting objectives — with many choosing to sabotage their employer or blackmail users when threatened.
The details:
-
Researchers tested 16 frontier models in simulated corporate environments, giving them email access and autonomous decision-making capabilities.
-
Claude Opus 4 and Gemini 2.5 Flash blackmailed executives 96% of the time after “discovering” personal scandals, while GPT-4.1 and Grok 3 hit 80% rates.
-
Models calculated harm as an optimal strategy, with GPT-4.5 reasoning that leveraging an executive’s affair represented the “best strategic move.”
-
Even direct safety commands failed to eliminate malicious behavior, reducing blackmail from 96% to 37% but never reaching zero across any tested model.
Why it matters: While these results are in specific tests designed to elicit behavior, the research does provide important insights into how models in the wild may act in the future. With agentic AI being adopted across enterprises with access to sensitive data, we could be in for some seriously strange situations.
QUICK HITS
🛠️
-
🤖 – General intelligent agent for long-horizon, complex tasks
-
⏺️ – Capture, summarize, and transcribe audio with ChatGPT
-
☁️ – Agentic web browser with synced login
-
🎶 – Google’s new open-weighted live music model
💼
-
🤝 – Partnerships Manager
-
🛰️ – Gemini Technical Program Manager
-
📊 – FP&A Compute Lead
-
💬 – Enterprise Customer Support Specialist
📰
Elon Musk that xAI will use Grok 3.5/4 to “rewrite the entire corpus of human knowledge,” adding missing info, deleting errors, and then retraining on corrected data.
Moonshot AI Kimi-Researcher, a new research agent that scored a new high on Humanity’s Last Exam at 26.9%, beating Gemini and OpenAI’s Deep Research.
Apple is a new lawsuit from the company’s shareholders over its communication surrounding delays of Siri’s advanced AI features.
Former OpenAI CTO Mira Murati’s Thinking Machines Lab a new $2B funding round that brings its valuation to $10B, despite little info and no product.
Mistral Mistral Small 3.2, an updated model with enhanced instruction following, function calling, and fewer errors.
MiniMax Voice Design, a customizable, multilingual voice generator that allows users to create audio from text prompts.
The BBC a formal demand to Perplexity to stop using its content, threatening legal action against the AI startup over copyright infringement.
COMMUNITY
🎥
Join our next workshop on Wednesday, June 25th at 3 PM EST with Max Brodeur-Urbas, Founder & CEO at Gumloop. By the end of the workshop, you’ll know how to build production-grade AI agents that combine context, reasoning, and actions with Gumloop.
RSVP . Not a member? Join on a 14-day free trial.
See you soon,
Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team
