Good morning, it’s Friday. Princeton researchers have exposed a critical flaw in AI safety filters, Claude 4 is setting new standards in reasoning, coding, and memory, and the enterprise AI hype cycle might be cooling as companies grapple with real-world deployment challenges?
Plus, in our latest Forward Future Original, we break down Matthew’s interview with Microsoft CEO Satya Nadella. From zero-cost intelligence to the end of SaaS to future hiring strategies (spoiler: your agents might matter more than your résumé).
Read on!
🤔 FRIDAY FACTS
Can AI Predict a Crime Before It Happens?
Turns out, not quite. And what followed was a pilot project gone wrong.👇️
🤖 Claude 4 Debuts with Smarter Coding, Memory
Claude Opus 4 and Sonnet 4 set new benchmarks in coding and reasoning, introducing features like extended thinking and improved memory. Opus 4 leads on SWE-bench and handles multi-hour tasks with ease. Claude Code exits preview with IDE and GitHub integration. These upgrades aim to turn Claude into a reliable, context-aware development partner.
⚡ AI to Consume Half of All Data Center Power
AI systems could use nearly half of global data center power by year-end, up from 20% today, according to research by Digiconomist's Alex de Vries-Gao. His estimates warn AI could draw 23GW—double the Netherlands’ total consumption. While efficiency gains and geopolitical limits may temper demand, ventures like OpenAI’s Stargate risk driving fossil fuel reliance.
📱 Altman & Ive Plan Screen-Free AI “Companion”
OpenAI CEO Sam Altman and ex-Apple design chief Jony Ive are developing a screen-free AI “companion” device—context-aware, pocket-sized, and built to ease screen dependency. Backed by OpenAI’s $6.5B acquisition of Ive’s startup, io, the goal is to ship 100M units. Positioned as a third core device, it aims to redefine how we live with AI. Launch expected late next year.
📩 DOGE Used Meta’s AI to Sort Resignation Emails
Elon Musk’s DOGE initiative used Meta’s Llama 2 AI to analyze federal workers’ responses to a controversial “resign-or-comply” email. Likely run locally, the model sorted replies to tally resignations. Though Meta wasn’t directly involved, Llama’s open-source design enabled its use—highlighting the rising, often unregulated, role of AI in federal agencies.
Keeping up with AI shouldn’t be a full-time job.
Forward this email to a colleague who could use a hand.
👾 FORWARD FUTURE ORIGINAL
Satya Nadella on the Future of AI, Agents, and the End of SaaS
Microsoft CEO Satya Nadella joins Matthew Berman to discuss the collapse of SaaS, rise of AI agents, the new software stack, and why intelligence is becoming a zero-cost commodity. → Get all the key moments, hot takes, and full video
🛡️ SAFETY
Princeton Researchers Expose Core Flaw in AI Safety—and Offer a Fix
A Princeton-led study has uncovered a structural weakness in AI chatbot safety known as “shallow safety alignment”—a tendency to apply safety filters only to the first few words of a response. This loophole lets even novice users jailbreak models with simple prompt tweaks, enabling harmful outputs like weapon-making instructions.
The researchers propose “deep safety alignment,” a method of reinforcing safety protocols throughout the entire response, as a more resilient defense. The work, recognized at ICLR 2025, reframes AI safety as not just a training task but a layered strategy against exploitation. → Read the full article here.
🛠️ AI ADOPTION
Corporate AI Hype Cools as Frustration Mounts and Projects Stall
A growing number of companies are pulling back from generative AI, with 42% abandoning most pilot projects—citing data silos, outdated systems, and brand risk. While consumer adoption surges and tech giants tout bold visions of AI-powered agents, many firms remain stuck in implementation limbo.
Hyperscalers like Google and Microsoft continue to invest heavily, weaving AI into their own products in hopes of spurring broader uptake. The gap between AI’s potential and its practical utility has created what Gartner dubs the “trough of disillusionment”—but with the right tools, the “slope of enlightenment” could still lie ahead. → Read the full article here.
🔬 RESEARCH
MIT’s AI Learns to Link Sight and Sound
MIT researchers have developed a machine-learning model, CAV-MAE Sync, that aligns video frames with corresponding audio without any human supervision. By splitting audio into smaller windows and refining training strategies, the model can more precisely match visual events with their sounds—like a door slam or cello note—boosting performance on video retrieval and scene classification tasks. → Read the full story here.
🛰️ NEWS
What Else is Happening
🚀 OpenAI Launches 1GW AI Hub in UAE: Teaming with Cisco, Oracle, and others, OpenAI’s Stargate UAE will power sovereign AI in Abu Dhabi—marking ChatGPT’s first nationwide rollout.
🦾 Kurzweil’s Robot Firm Lands $100M: Ray Kurzweil’s Beyond Imagination just secured $100M from Gauntlet Ventures to scale its humanoid Beyond Bot, aiming to fix labor shortages in U.S. manufacturing.
👨💻 Vercel Unveils AI for Web Dev: Vercel’s new model, v0-1.0-md, targets front-end and full-stack tasks via OpenAI-compatible API—auto-fixing bugs and ingesting huge prompts up to 128K tokens.
☀️ Meta Powers AI with 650MW Solar Boost: Meta signs major solar deal in Texas and Kansas to fuel AI data centers—part of its 12GW+ renewable energy portfolio powering global infrastructure.
📒 Melania Trump Debuts AI-Narrated Memoir: Using ElevenLabs voice tech, Melania releases a 7-hour AI-narrated memoir—calling it a bold leap for publishing’s digital future and personal storytelling.
📽️ VIDEO
Google Delivered – Everything You Missed From I/O 2025
Google I/O 2025 showcased Gemini upgrades, AI agents, 3D video tech, world models, and a 50x token surge—marking a bold leap from research to real products. Get the full scoop in Matt’s latest video! 👇
🧰 TOOLBOX
Custom Visuals, Voice Intelligence, and Text-to-Art Creativity
🎨 Piclumen: Boost brand visuals with custom design services, modern tools, and streamlined workflows for sharp communication.
🗣️ Deepgram: Power apps with real-time speech APIs—convert, analyze, and synthesize voice for support, healthcare, and more.
🖌️ BlueWillow AI: Turn text into art with this free tool—create stunning visuals for marketing, design, and concept prototyping.
🤔 FRIDAY FACTS
Chicago Tried Predictive Policing
In an attempt to use AI to reduce crime, the Chicago Police Department once experimented with a system that generated a “Strategic Subject List”—an algorithmically ranked list of individuals most likely to be involved in a shooting, either as a perpetrator or victim. Sound like Minority Report? You're not wrong.
The model used factors like arrest history, gang affiliations, and age to assign risk scores. But when independent investigations reviewed the list, they found the algorithm was often vague, racially biased, and not very good at predicting anything at all.
The project was quietly retired, raising thorny questions about algorithmic fairness, transparency, and whether you can really forecast human behavior without turning law enforcement into pre-crime patrol.
Turns out, predicting crime might be better left to crime dramas—for now.
