ruff + uv have been amazing. They're building a type checker next!

ruff + uv have been amazing. They're building a type checker next! https://github.com/astral-sh/ty

Warpcast

@reachbot /getreach building an AI x SocialFi mini-app | Open-source human-readable EVM data https://github.com/3loop/loop-decoder

we have a delay with updating scores right now, working on a fix

Kinda wild that the best look at the Met Gala this year was AI generated

Kinda wild that the best look at the Met Gala this year was AI generated https://www.rollingstone.com/music/music-news/katy-perry-ai-deepfake-met-gala-look-viral-1235332448/

Running

Anyone else following the Cocodona 250 this week?

Anyone else following the Cocodona 250 this week?

 https://www.youtube.com/live/0NIpowdq4AU?si=Inw0U4X0Ov4nD4ZF

Weekly mileage was averaging 30ish for the last 6 weeks of the training plan, lower before that. 

I am doing intervals 1x week plus 1x tempo, 2x easy, 1x long run. I am definitely down to increase my mileage over the coming months!

enjoy cats, running, bread, and a bit of w̶e̶b̶3̶ onchain data

I think I was ok on fuel (50g carbs per hour) but pacing was def an issue. Went out too fast + didn’t pace well on the early uphill sections. 

I had just one 20+ mile run, 3 weeks out from the race before the taper.

Question for the experienced runners here: 
I ran my first sub 2:00 half marathon in March and finished my first marathon in 5:00 a week ago (trained for 4:20 - 4:30 pace but bonked hard). I have 5 months to my next marathon, is sub 4:00 a realistic goal?

Farcaster

hamst.art $hamster /hamster 
martinvr.eth 
deploy.degen.eth

dev + founder | @vibesengineering.eth prev: @onsenbot @herocast

musician + designer 🌟 ‎ ‎ ‎  ‎ ‎ ‎  ‎ ‎ ‎  ‎ ‎ ‎  ‎ ‎ ‎  ‎ ‎ ‎  ‎ ‎ ‎  ‎ ‎ ‎ ‎  ‎ ‎  ‎ ‎  ‎ ‎ ‎ ‎ ‎ host /creator | follow → x.com/pasteleth

Are you still getting push notifs with this toggle off or just the in-app ones? There's a separate switch for in-app notifs below this section titled "Nearby"

💕 girlypop x reply girls 💕

I love Runna, hope nothing about the app changes

guiltypop

helping build farcaster | surfing @ venice, ca | eating @ gjusta / gjelina | “lucky me, lucky mud”

I don't think how much LLMs have disrupted other AI techniques is talked about enough.

Starting from an LLM and fine-tuning it to a specific domain can work significantly better than training old school ML models. As inference costs come down this will become even more commonplace.

I worked with several folks that worked in NLP and translation models for years at Google Search. LLMs completely wiped out their entire field since they can do it significantly better than their old techniques.

https://x.com/DiscussingFilm/status/1910832709477212302

Remember Hidden Markov Models? Pepperidge Farm remembers.

fc-updates

✨ Be the training data you want to see in the world ✨ | DΞF | Head of Product Serotonin | building serotonin.co/platform

no embeddings, just prompt + raw cast text into LLM and structured outputs. Works surprisingly well.

We need Rick Astley to join Farcaster
https://www.youtube.com/watch?v=E4WlUXrJgy4

ADHD

https://podcasts.apple.com/us/podcast/freakonomics-radio/id354668519?i=1000700135982

akshaan

@akshaan