CEO @entolabs, AI x energy engineering. Prev co-founder @wiseio acquired by @GE and machine learning x astrophysics @UCBerkeley
Denmark
Joined April 2009
- Tweets808
- Following509
- Followers611
- Likes1.4K
I used to think this but then I realized MCPs are much more effective if they are designed kind of like a CLI (and not like a REST APIs) but compared to CLIs you get:
- Nothing to install, just insert MCP URL
- No distribution or version hell, updates are available for users and agents whenever you deploy. Especially useful for new or fast moving servers of course
- OAuth built in, just hit connect
- With MCP apps you can embed UI elements as well, like charts and approval flows that the agent shouldn't see etc
So this chart from the Fable 5.1 release basically tells us to use Grok?
Today, we're releasing Fable 5.1 and Mythos 5.1. While the model is getting slightly better at a lot of things, I expect many users to appreciate that Fable 5.1 solves agentic coding tasks at roughly half the cost of Fable 5.
I personally also really like it's writing style, which imho sounds a lot more natural then our previous models did.
Some other personal highlights below!
Opened preferred coding agent, asked for a few tweaks to keybindings, created a git repo.
Now I can easily get any Omarchy env to run exactly the way I want 💪
This is how Linux will win! Congrats @dhh and team on the release and the foundation 👏
github.com/brinkar/omarchy-c…
How is it that NVDA is down so much when clearly everyone is rushing to buy more GPUs to run Deepseek? 😅
Damage control 😅 xitter.cf/StockMKTNewz/status/18…
Nvidia $NVDA just released a statement regarding DeepSeek:
"DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling. DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely-available models and compute that is fully export control compliant. Inference requires significant numbers of NVIDIA GPUs and high-performance networking. We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling."
"If an electron were the size of Earth, the experiment could detect a bump the size of a sugar molecule." 🤯
quantamagazine.org/the-elect…
Henrik Brink retweeted
Hey, ChatGPT-4:
stage 1. Decide on 16 random things: concepts, things, people, works of art, etc. make them as different as possible. Make them uncommon things
stage 2: set up a sweet 16 bracket between these things.
stage 3. Decide who wins each matchup in an entertaining way.
And multi-modal! I guess "I'm not a robot" CAPTCHAs will have to do something very different now? #gpt4
If not, it can always just hire and convince a person to do it 😂 xitter.cf/YosarianTwo/status/163…
Crossing my fingers for a successful deployment of #JWST at L2🤞 looking good so far!
I remember studying this project a decade ago 🤯 Can't wait for the amazing science that this telescope will enable 🔭
Carl Sagan testifies before Congress on climate change.
In 1985.
And here I thought I couldn't be a bigger fan.
Too bad they didn't listen, though...
m.youtube.com/watch?v=Wp-WiN…
