AI is leaving the screen

PLUS: Raycast launches Glaze to let users build Mac apps with AI

Author

Barsee
05 Mar

Together with

Howdy, it’s Barsee again.

Happy Thursday, AI family, and welcome back to AI Valley.

Today’s climb through the Valley reveals:

  • Raycast launches Glaze to let users build Mac apps with AI

  • AI startups are reaching $100M revenue at record speed

  • OpenAI is reportedly building its own GitHub-style platform

  • AI is leaving the screen

  • Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

NEBIUS

Courtesy: Nebius

Nebius Token Factory is designed for teams running open-source LLMs in real products. It provides fully managed inference with direct control over how requests are executed, including speculative decoding, cache-aware routing, and post-training tuned to real user traffic.

The result: consistent tail latency, predictable costs, and infrastructure built for AI systems that are already in production.

*This is sponsored

THROUGH THE VALLEY

Courtesy: Raycast

Raycast has introduced Glaze in private beta, a tool that lets Mac users build desktop apps by chatting with AI. Instead of generating web apps, Glaze creates fully native Mac applications that run offline and support features like keyboard shortcuts, menu bar tools, and background tasks. Users can keep improving their apps simply by describing changes in plain language. The beta is currently limited to Mac and early access is focused on Raycast users. Glaze also includes public and private app stores for sharing apps. Raycast says its experience building Mac productivity tools shaped the platform’s focus on native performance and user control.

Several AI products are scaling faster than most software companies in the past. Claude Code reached about $2.5B ARR just six months after launching in May 2025. Cursor, launched in March 2023, crossed $500M ARR in about 20 months. Lovable reached $200M within eight months of its 2024 launch, while Devin passed $150M in about 15 months. Emergent and Base44 both hit $100M in roughly eight to nine months, Bolt crossed the same mark in about a year, and Replit reached $100M about nine months after shifting to AI agents in 2024. In the past, even major software companies like Slack or Twilio often took five to ten years to reach $100M ARR.

OpenAI is developing an internal code platform that could replace Microsoft’s GitHub, according to The Information. The effort began after repeated GitHub outages during a large infrastructure migration to Azure frustrated OpenAI engineers. GitHub leaders told staff the full migration could take around two years, with several development teams reassigned to work on it. OpenAI is also considering opening the platform to external customers and integrating it with its Codex coding agents. If that happens, OpenAI would end up competing directly with GitHub, one of Microsoft’s most important developer platforms.

PEAK OF THE DAY

AI is leaving the screen

Courtesy: Oho.co

Artificial intelligence is starting to move beyond software and into the physical world. This shift is often called physical or embodied AI. Instead of only analyzing data or generating text, AI is being built into machines that can see, move, and interact with real environments. Examples include humanoid robots, self-driving cars, drones, and factory automation systems. Analysts expect this market to grow quickly, reaching between $500 billion and $1.4 trillion by 2035.

Building AI for the physical world is harder than software AI: Software models work with text and images, but physical AI must read sensors and react instantly in real environments while dealing with movement and safety risks. A chatbot giving a wrong answer is minor, but a mistake from a robot or self-driving car could cause real damage. Because of this, physical AI depends on strong vision systems, reinforcement learning, and fast real-time processing.

Autonomous vehicles are one of the biggest drivers of the sector and could generate up to $550 billion by 2035. At the same time, humanoid robots are moving from research labs into real workplaces. Companies like Unitree, Boston Dynamics, Figure AI, and Tesla are pushing robots toward practical use. Unitree plans to ship 10,000 to 20,000 of its G1 robots in 2026, starting at around $13,500, while Tesla aims to scale production of its Optimus robot with a long-term price target between $20,000 and $30,000.

Hardware is also improving: Robotics chips such as NVIDIA’s Jetson Thor allow robots to run AI models directly on the device instead of relying on cloud servers, helping them respond faster and operate more safely.

Physical AI is also expanding into other areas: Drones inspect infrastructure, agricultural machines remove weeds using lasers instead of chemicals, and robotic systems are assisting surgeons in healthcare.

AI is entering a new stage: For years, it existed mostly as software. Now it is becoming part of machines that can perform real work in the physical world, and this could reshape industries like transportation, manufacturing, healthcare, and agriculture over the next decade.

TRENDING TOOLS

SIGNALS

AI is advancing quickly, but the industry still faces real risks. Progress in new models could slow, which would turn competition into price wars while development costs remain extremely high. Chip shortages could also limit training and raise computing costs if demand outpaces global chip supply. Another risk is model dominance. If one lab’s system becomes far stronger than the rest, it could capture most revenue and investment, leaving smaller labs unable to compete.

Funding is another uncertainty. Investors are pouring billions into AI today, but that interest could fade if profits take too long to appear. At the same time, cheaper open models are improving fast and could reduce demand for expensive AI subscriptions. Governments could also introduce rules around energy use, safety, or data that raise costs or slow deployment. None of these risks guarantee a slowdown, but they show the AI boom is not without challenges.

WHAT I'M CONSUMING

THE VALLEY GEMS

What’s trending on social today:

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.