Anthropic just dropped what might be one of the most practical AI updates of the year, and it almost flew under the radar. I was catching up on the latest AI news when I saw this awesome video from an industry pro. The creator, Matthew Berman, was breaking down a bunch of recent headlines, but his analysis of a new Claude feature really caught my attention.
🧠 Claude’s New Brain: “Skills”
This innovator explained that Anthropic has launched “Claude Skills,” a way to give the model specialized knowledge that it can pull up on demand. It’s kind of like in The Matrix when Neo instantly learns kung fu. You can package instructions, code, and resources into a “skill” that Claude can access whenever it’s relevant.
The best part? It doesn’t bloat your context window. Claude is smart enough to only load the skill it needs for a specific task, which is incredibly efficient.
For example, the person who shared it showed how you could create a “Brand Guidelines” skill. You just put your style guide, logos, and specific instructions into a folder, upload it, and now Claude can help you write a marketing pitch that’s perfectly on-brand without you having to paste the guidelines in every single time.
⚡️ More Awesome Finds from the Video
The video was packed with other cool updates, too. Here are a few that stood out to me:
- NVIDIA’s Tiny Supercomputer: The expert highlighted that NVIDIA’s CEO is personally delivering the new DGX Spark, a “supercomputer in a box”, to top AI labs like OpenAI. The amount of power they packed into that small device is just wild!
- AI Discovering Science: I was blown away by this one. Google developed an AI model that generated a completely new hypothesis about cancer cell behavior, which scientists then went on to experimentally validate. This is exactly the kind of progress I get so excited about.
- Defining AGI: A new paper attempts to finally put a solid definition on AGI. The mind behind it proposed a framework that rates models on things like reasoning, memory, and speed. The paper’s early estimate puts GPT-5 at 58% of the way to their definition of AGI.
The original poster also covered everything from a hilarious Waymo “denial-of-service” attack in San Francisco to Jeff Bezos’s plan for data centers in space. It’s a fantastic roundup of what’s happening right now. You have to check out the full video for the complete scoop.