🔑

📞 🔥 🚬 🐼 it’s a new year

90s terminals hit different

@SherryYanJiang @DamiDina cursor, hermes agent, comfy ui

what's one ai tool that you think is criminally underrated??

@mycoliza few

i have spent my entire life waiting to receive this particular text message

@max_paperclips few

that's nothing. contemplate how people used to organise massive parties and raves, or meetup with friends, with no phones except their mums landline, and did this effortlessly

My son on why Caleb is his best friend at kindergarten: "He doesn't really speak English, so we can skip all the talking and just get right to the karate."

@snwy_me lmao

what the actual fuck is he talking about

@ns123abc Rate limits can get your account locked, beware

thinking of asking claude to make a script to automate unfollowing everyone i follow and restart bot free 🤔 wdyt chat?

highlander was about autism

you have no idea what’s going through someone’s single airpod and which ear matters ty

Which local models can actually handle tool calling? I built a framework to find out. 15 scenarios. 12 tools. Mocked responses. Temperature 0. No cherry-picking. Tested every Qwen3.5 size from 0.8B to 397B, and since some of you asked after the distillation tests: yes, I included Jackrong's Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled too. Only two models went all green: the 27B dense and the distilled 27B. The 397B? Failed two tests. The 122B? Failed one. The 35B? Failed two. The timed-out results — mostly on the smaller models, are cases where the model got stuck in a loop, repeating the same tool call until it hit the 30-second limit. The test that exposed the most models: "Search for Iceland's population, then calculate 2% of it." Simple, but 35B, 122B, and 397B all used a rounded number from memory instead of the actual search result. They didn't trust their own tool output. Small models hallucinate data. Big models ignore data. The 27B just threaded it through.

@keyOfIsis they can keep neptune

@kennfth_ Age of Aquarius reference

a lot of you are not prepared for a water world scenario and it shows

@PurzBeats @maxcapacity vibes

Claude got busy simulating the SNES PPU at the hardware level then corrupting it just to give lil' old me some nostalgic SNES glitches for my CRT display.

@CJZares @DanielleFong the archiving of posts will continue until morale improves (we retire)

@DanielleFong She’s goated. I have a bunch of her posts saved and whenever I’m bored I just go back and rewatch them to remind me how little I’m doing outside of work 😂

Kat is always doing some of the hottest shit

@DanielleFong literal cinema every post

@TheAhmadOsman we have ai on our own hardware at home

Just spent a couple hours playing with Hermes Agent (MiniMax M2.5 on a 2× RTX PRO 6000 node) Genuinely impressive experience MiniMax M2.7 weights will be the closest we’ve ever gotten to a fully local “Claude Code + Opus 4.6” experience Running on your own hardware at home

announcement coming

me saving links from the tl to my secret group chat of alts