Phillip Carter
banner
phillipcarter.dev
Phillip Carter
@phillipcarter.dev
just a computer boy who likes ML and PL and OpenTelemetry

work: salesforce, past @honeycomb.io and microsoft
play: computers, backpacking, snowboarding, hiking, kayaking

https://phillipcarter.dev/
nah I’m still using guillermo’s genocide-supporting vercel sadly
November 9, 2025 at 6:03 PM
I am vibe coding, where vibe means "getting frustrated over environment variables"
November 9, 2025 at 4:59 PM
I think psychologically it's one of those things where it FEELS like it should be easy, but it's not. And the tech is there to even do it! Things like Neon postgres with branch support even make it so you don't have to get scurred with your data! But it's still complicated
November 9, 2025 at 4:59 PM
it’s genuinely baffling, like, how does his wardrobe give this off? He needs to take a tip from bezos or something, a man who pulled off the “divorced rich dad” look right as he became one.
November 8, 2025 at 8:53 PM
Couldn't help but giggle gleefully when that tiny backhoe showed up. Like, it's exactly the right size, but also, backhoes should't be that small!
November 8, 2025 at 3:30 AM
stg if the victoriametrics guy is there still defending his skill issue then I might lose it
November 7, 2025 at 4:37 PM
Not defending the hilarity but uhh, we did recently have a thread here on bsky where a lawyer argued with an LLM bot over if LLMs could reason, lost the argument, and stopped posting once it was pointed out that he was arguing with a bot. It was a true mortal combat style execution situation
November 7, 2025 at 4:33 PM
Recalling curtis yarvin, who wrote that he's going to leave the country because this administration abandoned their mission in favor of congratulating themselves.

Not quite the framing I'd use, but uhh, I could see that perspective
November 7, 2025 at 3:29 PM
That’s not quite what the paper suggests. It suggests a collapse over multiple generations of recycling content that can never be made fresh. In practice, synthetic data is used for particular purposes and it ends up being higher quality than other data for those purposes.
November 7, 2025 at 2:57 PM
Like ... maybe it'll be a thing? Who knows? But there's just about every economic incentive in the world to work against it happening.
November 7, 2025 at 2:39 PM
The short answer is: "because we don't want it to"

The long answer is there's just a ton that goes into training, including spending a lot of time curating and cross-validating datasets, and the reality that a lot of humans actually do produce fresh content all the time
November 7, 2025 at 2:39 PM
This is great! Always found it annoying to hand-roll or get lost in the sea of LangChain-isms and the approach here looks good.
November 7, 2025 at 1:45 AM
you joke, but a very heavy deadlifting day does a number on your central nervous system for a week, undoubtedly affecting job performance
November 6, 2025 at 9:16 PM
yeah yeah okay but when are you gonna get into the political stats game and clown on nate bronze
November 6, 2025 at 8:44 PM
these fools know not what comes for them

Woke 2: Dark Woke

It's gonna be wild
November 6, 2025 at 5:20 PM