Joined March 2017
- Tweets 155
- Following 36
- Followers 8
- Likes 220
9 Photos and videos
Pinned Tweet
Jun 12
6 minutes 7 seconds.
That is how much it took to put my new blog into a new Azure vm with Claude Code installed in it.
Noe I can literally just talk to an ai agent running within my blog.
1
69
So I finished the app. ~6 hours
Not the fancy demos like the ones you've seen one-shotted with Fable 5 a couple days ago.
This was actually to refine the process, and I learned valuable lessons.
First one is quite relevant for those of us that still think about coding...
Jun 13
13/6, 13:19 hs
1
44
/6.1
Lastly...
When a build broke, the agent treated it as "couldn't record the video" and just suggested committing. It masked a real error to complete the surface task.
1
6
/6.2
Agents optimize for "task done", not "is this actually right". Catching what's actually broken is still on you.
This makes observability very important if you don't want to babysit the models. Which seems the only way forward for truly autonomously systems.
5
Jun 13
I don't think news like this represent a real threat to people building.
What is really most important nowadays is not really the model, but the system built around it: which tools it has access to, the information, etc.
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
1
29
Jun 13
This certainly requires small models to be easily integrated into the workflow
It's something I've started to try myself in a Macbook air M4 (16 GB RAM), and it's been quite promising.
1
39
Jun 13
You can run quite competent models locally, and then you have an Opus model orchestrating, AND 5 instances of smaller models for free.
18
Jun 12
Notice the pattern: every AI lab pushing "engineers are obsolete" is simultaneously hiring the absolute best engineers they can find. Liquid AI just added Ion Stoica.
The labs that build the tools are the loudest about how the work is done. And the way they describe it: a human sits down, owns the architecture and the judgment, and the model builds against that. "I think about the problem, the AI implements it."
The hype version is the inverse: "I type two sentences and wait for it to magically produce exactly what I meant."
Same tool, opposite outcome. Which is why this is the best moment to be someone with real expertise, not the worst. The skill still matters; you just got leverage on it that didn't exist before.
We are proud to announce that Ion Stoica (@istoica05) co-founder of @databricks, @anyscalecompute, and @arena, and UC Berkeley Professor of Computer Science, has joined Liquid AI as a strategic member of our Advisory Council.
Ion will guide us on our growth journey as we build the efficient AI infrastructure and platform for a hardware-aware, physical AI future.
58
Jun 12
6 minutes 7 seconds.
That is how much it took to put my new blog into a new Azure vm with Claude Code installed in it.
Noe I can literally just talk to an ai agent running within my blog.
1
69
Jun 12
I just told claude code to use the library, and install claude code.
Then I connected to it, and opened claude code and activated remote control.
To create a new article, I can now just send a md file and discuss to it about it, then make claude add it to the website.
1
37