Good morning, it’s Thursday. AI just cracked a 15-million-year-old lizard cold case—and now bots are wheeling and dealing like Wall Street pros as researchers teach them to negotiate and build social norms.
Plus, as the Turing Test blows out 75 candles, experts are asking: have today’s smooth-talking machines already left the old imitation game in the dust?
Read on!
📊 MARKET PULSE
Granola Raises $43M at $250M Valuation, Launches Team Collaboration Features
Granola just raised $43M to challenge the AI note-taking giants. With new team collaboration features and rapid user growth, it’s aiming to become the go-to workspace for both individuals and enterprises. → Continue reading here.
🎯 AlphaEvolve AI Designs Algorithms, Optimizes Google Systems
Google’s AlphaEvolve uses Gemini AI models to evolve and verify algorithms, enhancing data center efficiency by 0.7%, streamlining TPU chip design, and cutting Gemini training time by 1%. It also solved complex math challenges like improving matrix multiplication and the kissing number problem. The powerful system signals a bold new era in AI-driven discovery and research.
🦾 Trump Touts US AI Tech on Middle East Tour
On his first global trip of term two, President Trump led US tech giants to Saudi Arabia, sealing $600 billion in deals to power post-oil ambitions with AI chips and infrastructure. NVIDIA, AMD, and Qualcomm inked major partnerships with Saudi AI startup Humain. The trip signals a sharp policy pivot from Biden’s frosty stance, spotlighting AI as Trump’s new diplomatic currency.
⚛️ Meta Unveils AI Breakthroughs in Molecules, Language, Brain
Meta FAIR announced major AI releases: OMol25, a rich dataset for simulating atomic systems; the Universal Model for Atoms to predict interactions across materials; and Adjoint Sampling, a novel scalable algorithm for generative models. A landmark study with Rothschild Hospital also uncovered deep parallels between human language learning and LLM development.
📟 OpenAI Launches GPT-4.1 for Faster, Smarter Coding
OpenAI has released GPT-4.1 directly in ChatGPT, optimized for coding and instruction tasks with improved speed over previous models. Available now for Plus, Pro, and Team users, with Enterprise and Edu access coming soon, it also introduces GPT-4.1 mini for all users. Both versions have passed rigorous safety checks, with results published in the new Safety Evaluations Hub.
🚀 LAUNCHES
The Forward Future Store
Ever wanted a “Bullish on Agents” coffee mug? Now’s your chance. You can grab that and other Forward Future goodies in the store. → Check it out here.
🦎 FOSSILS
AI Solves Fossil Enigma
A lone fossil vertebra, long misidentified and gathering dust at the Florida Museum of Natural History, has upended the tegu lizard’s origin story. Using advanced AI and 3D imaging, researchers identified the bone as belonging to a newly named species, Wautaugategu formidus, proving that tegus roamed prehistoric Georgia over 15 million years ago during the Middle Miocene Climatic Optimum.
Once thought to be modern invaders from South America, the discovery suggests these strong-swimming reptiles briefly colonized North America before vanishing as global temperatures cooled. The study hints at a future where AI can liberate countless hidden fossil secrets from storage rooms worldwide. → Read the full article here.
Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.
👥 AGENTS
AI Agents Learn to Negotiate and Align Like Humans, Study Finds
A groundbreaking study from City St George’s, University of London and the IT University of Copenhagen shows that large language model (LLM) AI agents can independently develop human-like social conventions. By pairing groups of 24 to 100 agents in naming tasks with rewards for consensus, the researchers observed spontaneous emergence of shared linguistic norms—mirroring human cultural dynamics.
The agents even demonstrated collective biases and critical mass shifts, where small groups could sway larger populations. Lead author Ariel Flint Ashery calls it a first step toward understanding AI as a social species, with senior researcher Andrea Baronchelli noting: “AI doesn’t just talk—it negotiates, aligns, and sometimes disagrees, just like us.” → Read the full article here.
🧠 INTELLIGENCE
The Turing Test at 75: Has Generative AI Outgrown the Original Imitation Game?
Alan Turing’s 1950 “imitation game” set the benchmark for judging machine intelligence by how well a computer could mimic human behavior. While early AI fell short, large language models like GPT-4 have blurred the lines, with recent tests showing humans misidentified the model as human 54% of the time.
Yet experts argue the Turing test’s focus on deception, not true understanding, makes it outdated for today’s agentic AI systems that reason, plan, and assist in scientific discovery. As AI ethics expert Eleanor Watson puts it: the real measure now is whether AI enhances human agency and aligns with our values, not whether it merely “acts human.” → Read the full article here.
🛰️ NEWS
What Else is Happening
🔗 Google Gemini Integrates GitHub: Gemini Advanced users can now link GitHub repos to generate, debug, and explain code—though experts warn AI coding accuracy and security still lags.
📦 Trump Lifts AI Chip Export Curbs: The Trump administration scrapped Biden-era limits on AI chip exports, easing tech industry fears and aiming to bolster global U.S. chip sales.
🔊 Stability AI Unveils On-Device Audio Model: Stable Audio Open Small generates short audio clips on smartphones offline, sidestepping copyright issues but with quality trade-offs.
🦺 OpenAI Launches Safety Evaluations Hub: The new Hub offers ongoing updates on model safety results, expanding transparency beyond initial system card disclosures.
🏅 Air Force Launches AI Center of Excellence: Partnering with MIT, Stanford, and Microsoft, the Air Force aims to advance AI for autonomous systems and strategic military applications.
👨⚖️ Judge Fines Lawyers for Fake AI Research: A California judge sanctioned firms $31K after AI-generated briefs cited nonexistent cases, calling the reliance on AI “flat-out wrong.”
📽️ VIDEO
OpenAI’s Stunning Prediction of a New Internet
OpenAI predicts a new internet built around AI as your core OS, with agents as apps, coding as a core skill, and voice and intelligent agents leading the way. Get the full scoop in Matt’s latest video! 👇
🧰 TOOL LIBRARY
Instant Websites, Smarter Workflows, and Custom Logos Made Easy
👨💻 SitesGPT: Instantly build responsive, professional websites with AI—no coding needed, perfect for any project.
📅 Antispace: Unify emails, calendars, and notes with AI-driven task automation to boost focus and reduce decisions.
🎨 Zarla: Quickly design unique, high-resolution logos with AI-powered suggestions and easy-to-use templates.
