Good morning. It's Wednesday, March 25, and we're covering AI spotting landslide risks early, new multi-agent coding workflows, NASA’s plans for a permanent Moon presence, and more.
First time reading? Sign up here to keep up with the future of tech. If you enjoyed this email, share it with a friend.
NASA Plans Moon Base Push
NASA is building a scalable Moon base to support long-term human presence. Starting in 2027, it plans near-monthly landings of equipment, rovers, and scientific payloads. The effort includes investments in power, communications, and mobility systems. Robotic missions and drones will test surface operations ahead of human activity. The goal is to establish a sustained lunar foothold.
OpenAI Plans to Shut Down Sora App
OpenAI will discontinue the standalone Sora mobile app following mounting criticism over its realistic video-generation capabilities and privacy risks. The app, which allowed users to generate lifelike videos from text prompts, drew scrutiny for enabling deepfake-style content and “creepy” hyperreal outputs involving real people. OpenAI said it will instead fold Sora’s core technology into its broader platform, tightening safeguards and access controls.
Adcock Launches Personal AI Lab
Brett Adcock unveiled Hark, an AI lab focused on building highly personalized “personal intelligence.” He argues current chatbots feel limited and lack memory, perception, and real-world interaction. Hark aims to create systems that think like users and offload mental tasks. The company is also developing hardware to serve as a universal human-machine interface.
Alibaba Unveils AI Agent Chip
Alibaba introduced the XuanTie C950, a CPU designed to power AI agents and handle multi-step tasks. Unlike GPUs used for training, the chip targets inference workloads in data centers. Built on open RISC-V architecture, it offers customizable designs and claims 30% performance gains. The move supports China’s push for chip independence amid U.S. restrictions.
POWERED BY HERE.NOW
Free Instant Web Hosting for Agents
here.now is web hosting for agents:
Just tell your agent to publish to here.now
Your site is online instantly, no account needed
Perfect for anything you create with your agent and need a URL for: documents, websites, dashboards, tools, school work, visualizations, presentations, prototypes, games, articles, galleries, portfolios, media files, etc.
Works inside your favorite agents like Claude Code, Codex, Cursor and OpenClaw.
Free, no sign-up required.
REAL OR AI
Think You Can Still Spot AI?
Prove it. ↓ Real or fake. Answer below.
VIDEO
Hard Takeoff Has Started
THE FUTURE LIVE
ElizaOS Founder Joins the Live Show
“Feed-based social media won’t survive AI agents.”@shawmakesmagic Founder @elizaOS: “Platforms are already banning bots because they don’t know what to do.”“Almost every reply to everything I post on X is an agent now. Nikita is losing his mind over this. And I’m telling… pic.twitter.com/gCaP121gDM
@ForwardFutureView post on XGEOSCIENCE
AI Uses Satellite Radar to Predict Landslides Before Collapse
Researchers are using AI and satellite radar data to detect early signs of landslides—sometimes years before collapse—by measuring subtle ground movements invisible to humans. In Nepal’s Kimtang village, an AI-generated map identified a high-risk zone directly beneath homes, enabling authorities to plan evacuations and monitoring.
Similar systems analyzing Sentinel-1 satellite data have flagged 3,000 actively moving slopes across Great Britain and mapped thousands of landslides in disaster zones like Sumatra. The approach matters because landslides kill thousands annually and are increasing due to climate change and human activity. While not real-time or error-free, AI dramatically speeds analysis that would otherwise take humans months or years. → Read the full article here.
AGENTS
AI Coding Agents Improve by Splitting Tasks Into “Personas”
Developers are increasingly prompting AI coding tools to adopt multiple “personas”—such as product manager, engineer, and reviewer—to improve performance on complex tasks. This structured workflow breaks large jobs into smaller steps, helping models stay on track and produce more reliable code. Companies like xAI, OpenAI, and Anthropic are already incorporating versions of this approach through multi-agent or “subagent” systems.
Engineers report that this method reduces the need for detailed upfront instructions and improves end-to-end task execution. The shift highlights how better prompting strategies—not just better models—are unlocking stronger results from today’s AI systems. → Read the full article here. (Paywall)
NEWS
What Else is Happening
Huang Claims AGI, Then Clarifies Limits: NVIDIA CEO Jensen Huang said AGI has arrived, citing AI agents’ progress, but noted none can yet match human-level company building.
Arm Debuts AI Data Center Chip: Arm unveiled its AGI CPU, targeting fast-growing demand for agentic AI in data centers. The chip marks a strategic shift from licensing designs to building its own silicon.
Mirage Raises $75M for AI Video Tools: The Captions maker secured funding to build new video models and expand its marketing suite, betting on scalable content creation for creators, brands, and global markets.
Wozniak Questions AI’s Real Value: Apple co-founder Steve Wozniak says he rarely uses AI and finds it disappointing, arguing it lacks human nuance and shows no clear path to replacing people.
Doss Raises $55M for AI Inventory Layer: The startup builds AI tools that plug into ERP systems to sync inventory and accounting, targeting mid-market brands and positioning alongside AI-native finance platforms.
REAL OR AI
The Image Is AI. Tune in next week for another Real or AI test.
