Good afternoon here from Barcelona 👋🏽
It's been a while since my last newsletter, but I'm happy to restart the old “AI letters” - you can expect some updates in the coming editions, including changes to content, style, and how often I publish.
The line between generating video and programming robots is officially blurring. AI companies like Runway are now using the same "world models" that create video clips to build the ‘brains’ for physical automation.
This has huge implications for anyone running a physical business. Training a robot in a hyper-realistic simulation is exponentially cheaper and faster than real-world trial and error, opening up automation for logistics, agriculture, and small-scale operations.
Topics of the day:
Tesla’s bold gamble that humanoid robots will surpass its car business
China’s massive push into industrial automation
DeepSeek’s new agent that learns from experience
AI agents are gaining practical long-term memory
The Shortlist: Figure’s humanoid progress, chatbot reliability risks, and Orchard Robotics’ $22M raise
Tesla's All-In Bet on Humanoid Robots
The story: Elon Musk has predicted that humanoid robots will make up 80% of Tesla's future value, shifting focus from its multi-billion dollar car business to its Optimus robot initiative.
In practice:
The long-term vision is to automate physical tasks in warehouses, manufacturing, and logistics, tackling jobs that are dangerous or repetitive.
This could create a "Robots-as-a-Service" model where businesses lease robots for specific tasks, lowering the barrier to entry for advanced physical automation.
While robots aren't stocking your shelves yet, this signals a shift to start identifying which manual workflows you can digitize now in preparation for the next wave of automation.
Bottom line: This massive strategic bet signals that physical AI is moving from a research concept to a core business objective for major tech companies. It tells you where the next decade of R&D and venture capital will likely flow.
The Robotics Revolution Accelerates
What’s happening: China is installing nearly half the world’s industrial robots, around 280,000 units a year, to stay cost-competitive even as wages rise. At the same time, AI startups like Runway are turning video-generation “world models” into robot brains, using ultra-realistic simulations to train robots and self-driving cars.
In practice:
Robots can now be trained in millions of simulated scenarios, slashing costs and testing edge cases like sudden obstacles that are nearly impossible to stage in reality.
The same AI that generates marketing videos can simulate factory floors, logistics chains, or farm operations.
As these “brains” improve and drop in price, automation will spread from mega-factories into logistics, agriculture, and small business workflows.
Why it matters: The cost of intelligent, physical labor is dropping quickly. This unlocks opportunities to automate complex operational tasks that were previously too expensive or out of reach.
DeepSeek is building an AI Agent
What’s happening: Chinese AI firm DeepSeek is launching an AI agent by the end of the year, designed to rival OpenAI. The key difference is its ability to learn from experience while completing complex, multi-step tasks on your behalf.
In practice:
You could automate entire workflows, like having an agent conduct market research, analyze the findings, and draft a summary without step-by-step instructions.
These agents get smarter over time, learning from mistakes to optimize processes like lead qualification or customer support sequences with less human input.
Think of it as a digital employee you can assign to a goal, like “find 50 new prospects,” letting it handle the entire process from research to outreach.
Why it matters: This signals a shift from AI as a command-based tool to an autonomous partner. Soon, you’ll be delegating outcomes, not just tasks, freeing your team up to focus on strategy instead of execution.
AI Agents Get a Memory Upgrade
What’s happening: A new research framework called Memento is giving AI agents long-term memory, letting them learn from past successes and failures to improve future performance without expensive model retraining.
In practice:
You can build AI agents that get smarter over time, handling complex customer support tickets by recalling what solved similar problems in the past.
A research assistant could automate market analysis by remembering which data sources were most valuable for past reports, cutting down research time.
Automate sales outreach by having an agent learn from previously successful campaigns to personalize new email sequences for higher conversion rates.
Why it matters: Today's AI agents often start every task from scratch, limiting them to simple, repetitive work. This brings us closer to automated systems that learn and adapt like a human teammate, making advanced automation more practical for your business.
The Shortlist
Figure showed its 02 humanoid robot stacking and sorting dishes with high precision, demonstrating how its general-purpose AI model can learn new, complex physical tasks from training data alone.
NewsGuard found that major chatbots generate misinformation 35% of the time, a sharp increase from 18% last year, signaling growing reliability risks for businesses using them for content and research.
Orchard Robotics raised $22M to expand its vision AI technology for agriculture, which helps farmers more precisely monitor crop health and improve yields.
This newsletter is where I (Kwadwo) share products, articles, and links that I find useful and interesting, mostly around AI. I focus on tools and solutions that bring real value to people in everyday jobs, not just tech insiders.
Please share any feedback you have either in an answer or through the poll below 🙏🏽
