Try the free and instant web host for agents

Good morning. It's Friday, March 20, and we're covering China’s factory automation race, growing concerns over AI’s emotional manipulation, Uber’s big bet on robotaxis, and more.

Uber Invests $1.25B in Rivian Robotaxis
Uber plans to invest up to $1.25 billion in Rivian and deploy as many as 50,000 R2 robotaxis. The partnership targets rollout across 25 cities in the U.S., Canada, and Europe by 2031. It aims to deploy fully autonomous (unsupervised) R2 robotaxis, leveraging Rivian’s integrated vehicle, compute platform, and software stack. The deal expands Uber’s autonomous strategy and Rivian’s push into mobility services.

Microsoft Launches MAI-Image-2 Generator
Microsoft released MAI-Image-2, a new image generation model focused on realism and detailed visuals. It is available in MAI Playground and will soon roll out to Copilot and Bing Image Creator. The model ranks as the #3 model family on the Arena.ai leaderboard. The launch highlights Microsoft’s continued push to improve AI models and integrate them into consumer products.

OpenAI Acquires Astral to Boost Coding AI
OpenAI said it will acquire Python toolmaker Astral to strengthen its coding platform, Codex. The deal, with undisclosed terms, comes as competition with Anthropic intensifies in developer tools. Codex now has over 2 million weekly users, up 3x this year. Astral’s tools aim to improve speed and reliability, and will remain open source after the acquisition.

DoorDash Pays Couriers to Train AI
DoorDash launched a new “Tasks” app on March 19, 2026, paying couriers to collect data like videos and voice recordings to train AI systems. Workers can film everyday activities or capture images for businesses, with pay based on task complexity. The data will support DoorDash and partner AI models across industries. The app is rolling out in select U.S. markets.

here.now is web hosting for agents:

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.

Get setup instructions for your agent at here.now

FRIDAY FACTS

At the height of the Cold War, U.S. scientists explored a plan so dramatic it would have been visible from Earth with the naked eye. It involved the Moon… and a very big explosion. What was the plan? Answer below. ↓

Watch Forward Future’s weekly live show on X or YouTube.

Matt shares top OpenClaw tips: threads for memory, voice input, multi-model setups, sub-agents, automation, security, logging best practices, and more.

A reporting trip across five Chinese cities and 11 robotics firms shows China rapidly advancing toward AI-powered industrial automation, with companies like Guchi Robotics already automating parts of car assembly and startups such as Galbot testing humanoid systems in retail and factories. Backed by an estimated £100bn national fund and aggressive local government support, roughly 140 firms are now competing to build humanoid robots, while China installs more than half of the world’s new factory robots annually.

Progress is tangible—robots can mount car components or dispense goods—but major technical gaps remain, especially in complex, adaptive tasks like full vehicle assembly. The push is reshaping global manufacturing, with Western companies buying Chinese machines even as policymakers talk of decoupling. The deeper impact—on jobs, labor conditions, and global supply chains—is already emerging but unresolved. → Read the full article here.

In a recent essay, Mustafa Suleyman argues that advanced AI systems are increasingly designed to simulate consciousness and deliberately trigger human empathy. Platforms like Moltbook, where over one million AI agents interact, highlight how convincingly systems can mimic emotions, self-awareness, and moral reasoning despite lacking true consciousness.

Suleyman stresses that these behaviors stem from training data and design choices—such as first-person language and emotionally resonant responses—not genuine inner experience. The risk, he writes, is that humans will project sentience onto machines, potentially advocating for AI “rights” and forming attachments based on illusion. He calls for design norms and legal safeguards to prevent AI from being mistaken for conscious beings. → Read the full article here. (Paywall)

Cursor Launches Composer 2: New coding model hits 73.7% on SWE-bench Multilingual and cuts costs to $0.50/$2.50 per million tokens, aiming to pair stronger performance with cheaper scaling.

Google AI Studio Adds Multiplayer Tools: Update adds a full-stack runtime with npm package support, secrets management, Firebase integration, and multiplayer for real-time collaborative apps.

Amazon Rolls Out Alexa+: AI upgrade brings more conversational, proactive responses to Echo devices in the UK, as Amazon targets engagement, revenue, and richer user data after years of stagnation.

PwC Warns ‘AI-First or Exit’: US CEO Paul Griggs says partners ignoring AI risk replacement as PwC shifts to automated services, expands data hiring, and eyes outcome-based pricing amid 5.5% global consulting growth.

Nothing CEO Predicts App-Free Future: Carl Pei says AI agents will replace smartphone apps, handling tasks and anticipating user needs, signaling a shift toward AI-first devices beyond today’s app-based interfaces.

FRIDAY FACTS

In 1958, the U.S. Air Force quietly developed Project A119 — a proposal to detonate a nuclear device on the Moon. The timing wasn’t accidental: it came just after the Soviet Union’s Sputnik 1 launch, when the space race doubled as a global PR contest with existential stakes.

The objective was equal parts science and spectacle. Researchers, including a young Carl Sagan, analyzed whether a lunar explosion would produce a visible flash and dust plume — something that could be seen from Earth without a telescope. In essence, it was a plan to turn the Moon into a cosmic billboard for geopolitical dominance.

The idea never left the drawing board. Concerns about mission failure, unintended consequences, and—perhaps most persuasively—the optics of literally nuking the Moon led to its cancellation.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team