carloslemus.bsky.social
IoT and Edge Computing expert focusing on Energy and Industrial applications.
Systems Design/AWS/Python
LLMs because I like them, not because they're trending :-)
Indianapolis 🏡
27 posts
23 followers
52 following
Regular Contributor
Active Commenter
comment in response to
post
You can have the best of both worlds: lighting-fast local processing when you need it, and also cloud-scale intelligence when you want it.
comment in response to
post
Don't get me wrong. The cloud isn't going anywhere! But we're entering an era where the cloud handles the heavy lifting of analysis and insights, while edge devices can manage the immediate, critical operations—no matter how complex.
comment in response to
post
We're talking: complex AI models, real-time video analysis, and sophisticated sensor processing—all without the cloud's latency or
{data transfer costs}.
comment in response to
post
What changed? Two things:
1. Hardware has gotten incredibly powerful (and smaller!)
2. We've gotten better at building efficient applications.
**The result is that operations that previously needed cloud-scale computing can now run locally.**
comment in response to
post
Or Smart Manufacturing, where real-time quality control happens right on the factory floor using local AI.
comment in response to
post
Take EV Charging, my current endeavor, as an example: instead of sending every transaction all the way up to the cloud and back around, local processing can handle energy distribution operations within milliseconds.
comment in response to
post
It is *not* just a pendulum swing back to on-premise computing—it's way more interesting than that! Modern edge computing is truly expanding possibilities in how we can handle data-intensive operations.
comment in response to
post
And now I see that's what the original article starts with, which I only read after I read yours :-)
comment in response to
post
This reminds me of that trend last year where people were asking ChatGPT to make things increasingly "more X," as in, "more American"
comment in response to
post
Let's move tech forward honestly, reasonably, and peacefully.
Happy New Year!
comment in response to
post
(3) I will provide you with value-add content for the purpose of advancing my field only, not to trick the algorithm or steal your focus and attention.
comment in response to
post
(2) I will not further polarize the political spectrum. My opinions are my own, and I'm happy to admit I have too limited information or influence to sway most modern issues.
comment in response to
post
same carloslem.us
comment in response to
post
No, I made the false assumption based on your blog that the Flash 2.0 backend itself executed some code. It does not. But it can infer what the output *would* be for a given piece of code. TIL
comment in response to
post
No I think it's definitely using Google's massive dataset, because when I ask it to give me everything within the <head> tag for my site, it gets it *kind of* right, but it hallucinates:
comment in response to
post
Worked for my own personal site too, which I doubt was used for LLM training, but perhaps I'm wrong.
comment in response to
post
Worked again for Bluesky. It's either using knowledge of titles from its dataset, or it actually does have network access 🤔
comment in response to
post
Hi, the code with the requests library seems to have executed fine when I tried it using LibreChat.
comment in response to
post
I chuckled at that too :-)
comment in response to
post
Reasoning stuff but specifically for math and simple coding (not system design, though)