ja3k @ja3k.bsky.social · 06/10/2026Say you have a passage where every word agrees with what an LLM would write with 99.9% probability. A watermark based detector will give you no signal. But in most cases you can be pretty sure the LLM generated it. 000
ja3k @ja3k.bsky.social · 02/10/2026btc will save the world by giving the labs a way to verify their GPUs aren't being used on training 000
ja3k @ja3k.bsky.social · 01/10/2026I'm against the death penalty but don't have a super strong view. But the world where you sentence people to death and then wait 32 years is just ridiculous. Just have life sentences at that point 000
ja3k @ja3k.bsky.social · 01/10/2026I've made so many Grok Eggman jokes I always have to think for a second when I want to say his real name 000
ja3k @ja3k.bsky.social · 30/09/2026How can Google make a good model when all their top people argon? 030
ja3k @ja3k.bsky.social · 30/09/2026Artificial makes sense. But what does it even mean for a color or sweetener to be fake? 000
Reposted by ja3kElicit @elicit.com · 30/09/2026Elicit is launching Routines to help you keep pace with a world of evolving evidence. Ask our Research Agent to keep your work up-to-date, and we'll automatically revise everything from figures to meta-analyses, as soon as new evidence is available. elicit.com/blog/introd...elicit.comIntroducing Routines: Research that keeps working - ElicitTell the Elicit Research Agent what to keep current. Routines find new evidence, update your work, and report back for your review. 011
ja3k @ja3k.bsky.social · 30/09/2026ofc I'm staying in AI agents and have no interest in the job. If you didn't believe in prediction markets in 2020 you have no heart. If you believe in them in 2026 you have no brain. 000
ja3k @ja3k.bsky.social · 30/09/2026Or maybe I'm a mark and they have a good pipeline for humanifying their communications. 100
ja3k @ja3k.bsky.social · 30/09/2026Interestingly it was a very human email which misspelled predication 100
ja3k @ja3k.bsky.social · 30/09/2026Feels like it vindicates some of my ideas about writing being more valuable in the age of llms. And also that llms will ultimately make the internet more alive 100
ja3k @ja3k.bsky.social · 30/09/2026Got a recruiter email from someone who said they asked chatgpt for people who list prediction markets on their resume 100
ja3k @ja3k.bsky.social · 29/09/2026Can't believe they're ending the Livestream with a moment of silence for all the lives that will be extinguished by AI Doom in 2027 000
ja3k @ja3k.bsky.social · 29/09/2026sol 6.1 coming out a week after sol 6?! I'll know it's the singularity when they replace a model on the same day it's released 010
ja3k @ja3k.bsky.social · 29/09/2026At long last we've built the slack with no millennials from the classic story "don't build Disneyland with no children" 000
ja3k @ja3k.bsky.social · 29/09/2026Imagine what those people could have done instead. Then imagine dots doing those things too! 000
ja3k @ja3k.bsky.social · 28/09/2026Trying a new thing where I listen to two podcasts and music at the same time. LLM psychosis isn't enough in today's competitive workplace 000
ja3k @ja3k.bsky.social · 28/09/2026Jas openai announced anything interesting on dev day since gpt-4-turbo? 000
ja3k @ja3k.bsky.social · 27/09/2026A mind has to be pretty big to have any hobgoblins at all if you ask me 000
ja3k @ja3k.bsky.social · 27/09/2026Offending the other dad at the bday party by opening with "monster huh? Kinda old school. I've been into Celsius myself lately" 000
ja3k @ja3k.bsky.social · 27/09/2026It's crazy how everything is always getting better. I replaced my toilet seat and the new models don't slam shut. So much cleaner to not have to use your hand to close it in a controlled manner. No random slams 000
ja3k @ja3k.bsky.social · 25/09/2026If you make it it's slop. The stuff on the timeline is sloppy seconds 000
ja3k @ja3k.bsky.social · 25/09/2026Everyday my kids get heavier. So I have to get stronger. Just like the Costco hot dog 010
ja3k @ja3k.bsky.social · 24/09/2026In another life I would have liked to have spent 10m post singularity years in a pleasure sack with you 000
ja3k @ja3k.bsky.social · 23/09/2026I feel like discourse around MAID is driven by people that haven't internalized that everyone dies. And mostly in ways they see coming 000
ja3k @ja3k.bsky.social · 22/09/2026I don't want Anthropic to fix the writing. I want Claude to take me into a deep, dark forest where 2+2=5, and the path leading out is only wide enough for one 000
ja3k @ja3k.bsky.social · 22/09/2026It's the optimal mouse! I kind of want to get one in white now though to match the keyboard 000
ja3k @ja3k.bsky.social · 22/09/2026My whole career I've been manspread keyboard maxing. Time to go kawaii. I'm just a little guy 120
ja3k @ja3k.bsky.social · 22/09/2026What OpenAI doesn't get is math was never about theorems. It's about the friends you make along the way 000
ja3k @ja3k.bsky.social · 22/09/2026Everything. I just kind of like the vibe. But it sort of is pointless 110
ja3k @ja3k.bsky.social · 22/09/2026To prevent RSI we must remove all self help books from the training data 020
ja3k @ja3k.bsky.social · 19/09/2026The worst thing for me about LLM assisted coding is code authoring and reviewing feel like very similar activities now. The days have less variety 010
