Dave
@skillmcp
Followers
3K
Following
53K
Media
491
Statuses
5K
Claude app store soon
BREAKING 🚨: Anthropic is working on @ mention support for Claude Cowork, where you will be able to trigger certain MCP capabilities. However, some placeholders are named as "MCP Apps" and "Imagine" widgets for SVGs and HTML components. This may be a reference to a solution
0
0
2
Convex just got Claude Code support. 12 installable skills covering: - Best practices - Realtime patterns - Schema validation - File storage - HTTP actions - Migrations
2
0
3
Agent Skills are a new attack vector for bad actors. As skills gain popularity remember to read the original article on Skills from Anthropic:
3
0
3
The ONLY real difference between skills and MCPs is auth. If you don't need external auth, skill is simpler. If you need OAuth or API keys to external services, MCP makes sense.
1
1
5
MCP servers with 50+ tools each, preloaded into context whether you use them or not. Lazy loading was the obvious answer. Took a while but it's here.
0
0
2
I wouldn't be surprised if we see a 1M context window via Sonnet/Opus for general public in February. That would remove my model swapping to Gemini.
1
0
1
Anthropic goes white collar Google goes hardware OpenAI goes health xAI goes military
1
0
5
Everything I can do with Claude Coworker I've already been able to do with Claude Code. Great for mass adoption, not so much for anything else. The killer feature would be Claude app synced with my local computer so I can go away and still do things on my desktop or server.
1
0
3
So this is why they restricted OAuth from third party harnesses
In Cowork, you give Claude access to a folder on your computer. Claude can then read, edit, or create files in that folder. Try it to create a spreadsheet from a pile of screenshots, or produce a first draft from scattered notes.
0
0
3
if you own a top-tier Gen4/Gen5 NVMe right now you're ready for beta. But the hardware meant to run these Terabyte-scale models smoothly at a commercial scale hasn't even hit the shelves yet. https://t.co/0wdbUfG7Xy
github.com
Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models - deepseek-ai/Engram
0
0
0
High-IOPS NVMe SSDs will be the industry's cheapest VRAM. We are going to see SSD controllers specifically tuned to handle AI Texture Streaming effectively.
1
0
0