💸 Musk’s xAI Sells Grok to Feds for ¢42
xAI will offer its Grok 4 and Grok 4 Fast models to U.S. federal agencies for 42 cents per agency under an 18-month contract. That rate undercuts rivals: OpenAI and Anthropic charge roughly $1 per agency, while Google’s Gemini is priced at 47 cents.
📰 Microsoft Plans AI Marketplace to Pay Publishers
Microsoft is piloting a Publisher Content Marketplace to pay media outlets per use of their content in AI tools like Copilot. Unlike flat licensing deals, it aims to build a scalable, two-sided model. If successful, it could shift AI content economics and push rivals like Google to respond.
💰 Databricks Bets $100M on OpenAI for Enterprise AI
Databricks is embedding OpenAI’s models, including GPT-5, into its platform under a $100M deal to drive enterprise AI use. Customers can now build tailored agents on secure data via Agent Bricks. Databricks takes on revenue risk; OpenAI secures guaranteed income.
🖥️ Windows ML Lets AI Apps Run Locally on PCs
Microsoft has released Windows ML to all developers, enabling AI apps to run directly on Windows 11 hardware. It auto-selects the best CPU, GPU, or NPU for each task, reducing cloud dependency. Early adopters like Adobe and McAfee are using it to power real-time, on-device AI features.
🤔 FRIDAY FACTS
What Happens When AI Simulates Thousands of People?
Forget focus groups and survey panels, researchers are now asking AI to “play the role” of thousands of virtual participants in scientific studies. But can AI really replicate how real humans respond to personality tests, political questions, or moral dilemmas?
Stick around to find out! 👇
🪧 POWERED BY ROCKET
Turn Ideas into Apps - No Code Needed
Rocket, the vibe solutioning tool built for creators, builders, and AI-first teams. Describe your idea in plain English or drop in a Figma file and Rocket produces a complete web or mobile app frontend, workflows, backend logic, and launch-ready code.
You’ll also receive 1M free tokens to kick off your first build. Whether you’re validating a concept or scaling a product, Rocket accelerates your path from vision to value.
📽️ VIDEO
Open-Source Multi-Agent Framework
In this tutorial, Matt shows how to build real-time business reports using AWS Strands, a free, model-agnostic multi-agent framework.
🏢 WORK
AI in the Office: Quietly Transforming How We Work, Talk, and Get Judged
The Recap: This article from The Economist explores how generative AI is subtly but pervasively reshaping white-collar work environments. While not yet revolutionary, AI tools like ChatGPT and Claude are altering office behaviors, language, and norms—often without formal oversight. The story contrasts this quiet shift with the overhyped metaverse wave of the early 2020s, noting that AI’s impact is already far more tangible.
Highlights:
Work-related ChatGPT usage surged from 213 million messages per day in June 2024 to 716 million by June 2025, according to a study by OpenAI and Harvard’s David Deming.
Anthropic's data shows “automation” now surpasses “augmentation” in workplace AI use, signaling a shift toward task delegation rather than collaboration.
Employees often adopt AI tools independently or secretly, uncertain whether usage will be rewarded or penalized.
AI now transcribes meetings, tracks employee engagement, and monitors candidates during interviews, blending into routine corporate oversight.
Writing tasks dominate AI use at work, increasing fluency but also resulting in more bland, generic content.
Forward Future Takeaways:
AI’s integration into the office isn’t flashy—but it’s deepening. From automating tasks to influencing language and surveillance, generative AI is changing how people work, communicate, and are evaluated, often ahead of formal company policy. → Read the full article here. (Paywall)
🔬 SCIENCE
MIT’s MultiverSeg AI Slashes Time Needed to Analyze Medical Images
MIT researchers have unveiled MultiverSeg, an AI-powered tool that dramatically accelerates medical image segmentation—a foundational step in clinical research. Unlike traditional systems, MultiverSeg improves with use, requiring fewer clicks and no retraining, enabling scientists to annotate entire imaging datasets with minimal effort.
The model learns from previous inputs and uses a flexible context set to refine predictions, reaching high accuracy with just a few examples. By cutting hours of manual labor into minutes, this innovation could fast-track studies on treatments and disease progression, while reducing the cost and complexity of clinical trials. → Read the full article here.
🦾 ROBOTICS
Google Debuts Gemini Robotics-ER 1.5: A Brainy Boost for Smarter, Safer Robots
Google has released Gemini Robotics-ER 1.5, its first publicly available AI model designed specifically for embodied reasoning in robots. Optimized for tasks like sorting, planning, and spatial understanding, the model can interpret natural language, reason across time, and call external tools—including robot APIs and web search—to complete complex physical tasks.
Developers can now fine-tune the “thinking budget” to trade off speed and accuracy, while improved safety filters help the model avoid risky or impossible actions. By combining spatial precision with high-level task orchestration, Gemini ER 1.5 marks a significant step toward robots that can not only see and act, but understand. → Read the full paper here.
🛰️ NEWS
What Else is Happening
🥡 Curry-Backed AI Tackles Food Supply Chaos: Burnt, an AI startup led by seafood industry vets, raised $3.8M to automate messy food order systems with agents that work on top of legacy software.
🔥 GPT-5 Goes Pro: OpenAI’s latest benchmark shows GPT-5 rivals human experts in 40% of tasks across major industries, marking a big step toward real-world utility for advanced AI systems.
👨🔬 Meta Snags OpenAI Scientist (Paywall): Yang Song, who helped shape DALL·E 2 and led strategic research at OpenAI, has joined Meta’s AI lab as research principal—another major Zuckerberg hire.
📉 Oracle Plans 10K Layoffs Amid AI Pivot: Oracle may cut 10,000 jobs as it refocuses on cloud and AI. Legacy apps face downsizing, while agentic AI could let customers build features once done in-house.
🎵 Spotify Purges 75M AI Spam Tracks: Spotify removed 75M “spammy” songs and introduced stricter rules to combat AI abuse and vocal impersonations—aiming to protect real artists and fans.
🧰 TOOLBOX
Trending AI Tools
🤔 FRIDAY FACTS
AI Can Simulate Populations, And Their Answers Match Ours
In 2022, researchers discovered that large language models like GPT-3 and GPT-4 could simulate thousands of virtual participants in social science experiments—with surprisingly human results. When given psychology surveys, political questions, or moral dilemmas, the AI-generated responses reflected real-world population trends across age, gender, ideology, and even culture.
In one study, the models reproduced known differences between American and Japanese participants, just by being prompted with culturally relevant context. In another, simulated “populations” scored on the Big Five personality traits in distributions nearly identical to actual survey data.
That’s a Wrap!
📢 Want to advertise with Forward Future? Reach 550K+ AI enthusiasts, tech leaders, and decision-makers. Let’s talk—just reply to this email.
Thanks for reading today’s newsletter—see you next time!
Matthew Berman & The Forward Future Team
🧑🚀 🧑🚀 🧑🚀 🧑🚀
