🧑🏭 OpenAI Takes On LinkedIn With AI-Powered Jobs Platform
OpenAI just announced it’s launching the OpenAI Jobs Platform—a matchmaking service using AI to pair employers with talent—slated to go live by mid-2026. Led by Fidji Simo, the initiative includes a dedicated track for small businesses and local governments, broadening access beyond large enterprises. The company will also offer tiered OpenAI Academy certifications to validate AI skills and build trust in its candidate pool.
🚨 Hackers Exploit Grok AI to Spread Malware
Cybercriminals are abusing X’s Grok AI to bypass ad restrictions and distribute malware. By embedding links in metadata and prompting Grok to surface them, attackers gain massive visibility. The tactic, dubbed “Grokking,” boosts SEO and trust. Guardio Labs found hundreds of accounts using this method in a coordinated campaign.
📱 Google Launches On-Device Embedding Model Gemma
Google DeepMind unveiled EmbeddingGemma, a 308M parameter model built for on-device AI tasks like RAG and semantic search. It offers top-tier multilingual embeddings, works offline, and fits under 200MB RAM with quantization. Ranking first in its class on MTEB, it’s integrated with tools like llama.cpp and transformers.js for easy deployment.
🛒 Atlassian Acquires Arc Maker for $610M to Reinvent Browsing
Atlassian is buying The Browser Company—maker of Arc and Dia—for $610M in cash, aiming to build an AI-first browser tailored for work. CEO Mike Cannon-Brookes says the deal reimagines browsers for SaaS-heavy, tab-driven workflows. The startup will operate independently, continuing to develop Dia while scaling faster under Atlassian’s wing.
Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.
🤔 FRIDAY FACTS
How Much Faster and More Accurate Are AI-Powered Forecasting Systems?
Weather forecasting has always been a race against time—especially when lives and livelihoods are on the line. But in 2024, DeepMind introduced GenCast, a generative AI model that’s shaking up meteorology.
Stick around for the answer! 👇
🔍 SEARCH
Apple Bets on DIY AI Search While Wall Street Wants a $30B Perplexity Buy
Apple is reportedly developing its own AI-powered search tool, "World Knowledge Answers," as part of a rebuilt Siri set to debut by March 2026, sidestepping analyst calls to acquire Perplexity AI. Despite pressure to join the AI arms race through big-ticket deals, Apple is sticking to its asset-light, privacy-first strategy, emphasizing on-device processing and select partnerships.
Critics, like Dan Ives, argue the company is “watching from the stands” as rivals like Microsoft and NVIDIA lap it in AI innovation. But Apple’s $3T war chest and cautious M.O. suggest it’s still betting it can win the race, on its own terms. → Read the full article here.
🏠 SMARTHOME
SwitchBot Unveils AI Hub, Art Frame, and Emotional Pet Robots at IFA 2025
SwitchBot is expanding its quirky smart home lineup with three AI-driven devices: a next-gen AI Hub that interprets events like a fall or misplaced phone, a colorful E Ink Art Frame that generates custom artwork via prompts, and two penguin-like companion robots named Niko and Noa.
The AI Hub features on-device and cloud intelligence powered by a 6T chip, while the bots can recognize emotions and exhibit feelings like jealousy. Though pricing and release dates are TBD, the company’s solid hardware track record suggests these AI-infused gadgets are more than just show floor stunts. → Read the full article here.
🦾 ROBOTICS
ETH Zurich’s Robot Dog ANYmal Learns to Play Badminton — And It’s Surprisingly Agile
Researchers at ETH Zurich have trained a quadruped robot named ANYmal to rally in badminton matches against human players, showcasing up to 10-shot exchanges. The AI-powered robot combines whole-body motion with visual tracking to react and return shuttlecocks with surprising dexterity.
While ANYmal was originally designed for navigating rough terrain, this latest feat demonstrates the potential of four-legged robots in dynamic, real-time environments like sports. It's a leap forward in teaching machines to coordinate perception and movement under pressure, birdie included. → Read the full article here.
🤖 MODELS
Switzerland Releases Apertus, a Fully Open-Source AI Model Built for the Public Good
Swiss institutions have launched Apertus, a multilingual, fully open-source AI model developed by EPFL, ETH Zurich, and CSCS. Available in 8B and 70B parameter versions, Apertus is open down to its training data and architecture — a rarity among large language models. It was trained on 15 trillion tokens in over 1,000 languages and is already live on Swisscom’s sovereign AI platform.
Designed to serve as public infrastructure, Apertus aligns with EU AI Act standards and positions Switzerland as a leader in transparent, democratically governed AI development. → Read the full paper here.
🛰️ NEWS
What Else is Happening
👋 Turing AI Chief Resigns Over Defense Shift: Amid staff unrest and a government threat to pull funding, the UK’s top AI institute loses its CEO as it pivots from research to military priorities.
🎞️ Google Photos Adds Veo 3 for AI Videos: Users can now animate still images into short, high-quality clips with Veo 3 — Google’s latest move to put cutting-edge AI in everyday apps.
💰 Quantinuum Raises $600M at $10B Valuation: Honeywell’s quantum spinout secured fresh funding from NVIDIA and others, doubling its value and signaling investor appetite for next-gen computing power.
📉 California Leads U.S. in 2025 Job Cuts: Driven by AI, economic worries, and federal downsizing, layoffs have already topped 2024 totals — with California hit hardest, especially in tech.
📽️ VIDEO
AI News: XAI Sues OpenAI, Microsoft's MAI, Anthropic Funding, OpenAI Acquisition, and More!
Elon sues ex-XAI engineer over OpenAI leak, Microsoft launches AI models, Anthropic raises $13B, Google wins antitrust case.
🧰 TOOLBOX
Improved Feedback, Smarter AI Apps, and Deeper Market Research
🌟 CannyCollect and prioritize user feedback, manage feature requests, and share product roadmaps to build better tools. | 🛠️ DifyBuild and deploy custom AI apps with this open-source platform designed for seamless integration and scale. | 📊 TegusAccess expert calls, qualitative data, and market research to fuel smarter business and investment decisions. |
🤔 FRIDAY FACTS
Forecasting the Future with GenCast
GenCast isn’t your average weather model—it’s a diffusion-based AI system, borrowing the same tech behind image generators like DALL·E. But instead of rendering pixel-perfect cats, GenCast produces high-resolution simulations of future weather, one frame at a time.
Traditional forecasts rely on numerical weather prediction (NWP) models that solve complex physics equations—a process that can take hours on supercomputers. GenCast skips the physics and learns directly from decades of historical data, generating predictions using pattern recognition instead of raw math.
The result? GenCast produces global forecasts in minutes on a single TPU instead of hours on supercomputers, with comparable or better accuracy for 1–3 day forecasts.
🗒️ FEEDBACK
Help Us Get Better
What did you think of today's newsletter? |
