Charlie Francoise
@loderunner.bsky.social
"Don’t fire your SRE team tomorrow thinking that Claude will be on-call"
I have a feeling it’s going to happen anyway.
I have a feeling it’s going to happen anyway.
November 6, 2025 at 4:59 PM
"Don’t fire your SRE team tomorrow thinking that Claude will be on-call"
I have a feeling it’s going to happen anyway.
I have a feeling it’s going to happen anyway.
Slides are back up, and they look great. I love the Anthropic aesthetic.
November 6, 2025 at 4:39 PM
Slides are back up, and they look great. I love the Anthropic aesthetic.
I like the terminal-driven approach to constrained LLM tasks. Rémi really tying into @fabien.potencier.org & vaibhav’s talks.
November 6, 2025 at 4:00 PM
I like the terminal-driven approach to constrained LLM tasks. Rémi really tying into @fabien.potencier.org & vaibhav’s talks.
The talk about "reliability" would have landed better if the LLM calls hadn’t spewed out different outputs for the same input.
November 6, 2025 at 3:57 PM
The talk about "reliability" would have landed better if the LLM calls hadn’t spewed out different outputs for the same input.
"We’re building error-handling systems that occasionally do AI"
Big oof. that one hit close to home
Big oof. that one hit close to home
November 6, 2025 at 3:45 PM
"We’re building error-handling systems that occasionally do AI"
Big oof. that one hit close to home
Big oof. that one hit close to home
"UNIX tools cannot understand the meaning of text"
Neither can Claude, Rémi...
Neither can Claude, Rémi...
November 6, 2025 at 3:42 PM
"UNIX tools cannot understand the meaning of text"
Neither can Claude, Rémi...
Neither can Claude, Rémi...
he did talk about BAML in the end
November 6, 2025 at 12:23 PM
he did talk about BAML in the end
C’mon, Mirakl, support a designer
November 6, 2025 at 11:51 AM
C’mon, Mirakl, support a designer
"prompts are functions" is what I’ve been saying for a year now
November 6, 2025 at 11:10 AM
"prompts are functions" is what I’ve been saying for a year now
"If you don’t write test cases with probabilistic code, you’re in for a world of hurt" 😂
November 6, 2025 at 11:10 AM
"If you don’t write test cases with probabilistic code, you’re in for a world of hurt" 😂
I can actually get on board with this: We don’t need new engineering concepts just because LLMs. (Do we need new tools?)
November 6, 2025 at 11:10 AM
I can actually get on board with this: We don’t need new engineering concepts just because LLMs. (Do we need new tools?)
What I’m seeing is that, after Python had a minute in the spotlight while AI was an ML problem.
Now that LLMs have been commoditized, JS is coming back strong!
Now that LLMs have been commoditized, JS is coming back strong!
November 6, 2025 at 10:40 AM
What I’m seeing is that, after Python had a minute in the spotlight while AI was an ML problem.
Now that LLMs have been commoditized, JS is coming back strong!
Now that LLMs have been commoditized, JS is coming back strong!
Apparently, it means that:
* systems will be composed of multiple agents
* agents need a good model of their domain
* agents need a clear contract for communication
This is indeed how you build distributed systems.
* systems will be composed of multiple agents
* agents need a good model of their domain
* agents need a clear contract for communication
This is indeed how you build distributed systems.
November 6, 2025 at 10:40 AM
Apparently, it means that:
* systems will be composed of multiple agents
* agents need a good model of their domain
* agents need a clear contract for communication
This is indeed how you build distributed systems.
* systems will be composed of multiple agents
* agents need a good model of their domain
* agents need a clear contract for communication
This is indeed how you build distributed systems.
Reposted by Charlie Francoise
OpenAI did not ask for government support, unless I have missed something. Their cfo, who keeps saying stuff that no cfo should say, said some bizarre shit.
November 6, 2025 at 6:15 AM
OpenAI did not ask for government support, unless I have missed something. Their cfo, who keeps saying stuff that no cfo should say, said some bizarre shit.