kuna

as a data scientist/"AI engineer", i partially agree with you we DEFINITELY shouldn't be integrating AI into everything, but semantic understanding is a powerful function that can improve many areas of software your siri example from yesterday is prime, but the fact that an app intent requires certain phrasing is the biggest issue and was the point i was trying to make yesterday

What’s funny is a lot of people have given me crap in the comments defending this But I’ve gotten messages from AI and UX devs at almost every major company and lab agreeing that this isn’t great and AI UX needs work

@alexhldk yeah. i just wanted to know if quinn had other reasons like personality traits, age, professional history, etc.

@asikunaa Look at the current state of their software, macOS especially, and take a wild guess. It is as much on Federighi as it was on Dye.

This guy should’ve been Apple’s new CEO

@SnazzyLabs @doganuraldesign i agree but why not in your opinion

@doganuraldesign No.

@mweinbach the problem is that you need forethought to set this up beforehand, rather than it learning from your context and becoming an actual “assistant”

All I’m going to say is for all the crap people give Siri, and most of it is deserved, it is still really good at doing a lot of things The problem, and why I think that it’s still deserved, is if you try to do this too, there’s a good chance it won’t work

@mweinbach why parakeet v3 instead of v2? are you doing multi-lingual? v2 is better at english-only ASR

ok now at 180 rtfx, but quality of the model dropped ~3% because we quantized to 8-bit proper precision is 125 rtfx... maybe it's worth it? tbd, let's see what else we can do

We know you know, but how about we tell you some more now? Tune in for Assassin's Creed Black Flag Resynced's Worldwide Reveal Showcase on April 23rd at 4PM UTC - 6PM CEST - 9AM PDT. youtu.be/7hkFaTypP1Q

@assassinscreed OH MY FUCKING GOD

@serialdotai hey

I need to follow more ml engineers

@Mrwhosetheboss GEMINI????

Claude > Gemini > ChatGPT It's not even close right now

@Toribabieegirl trust yourself more

You meet your 16 year old self, you’re allowed 3 words. What will you say?

@MelkeyDev god no

Opus 4.7 > Codex 5.4? 🤔

@RihannaFohead @tsuakies what the fuck is the audio

@tsuakies Took so many Ls my body is just-

how it feels finally winning after a long loss streak

@inerati allo

@LirienWires @iShowShitpost valve doesn’t care brodie kids have publicly inherited their parents steam accounts before

@iShowShitpost Kinda sad about this because policies like no sharing or giving accounts to friends or family members are not allowed so the account gets banned or never used again

@Hondunu @Dexerto by scrolling in the comment section…?

@Dexerto 100 likes in less than an hour? How are y'all finding this?

Square Enix has created an AI tool that could cut up to 3000 hours from manga production It automates typesetting like font, size, and placement

@garrytan the last time openclaw was actually functional was when it was still called clawdbot

Using OpenClaw is basically is like driving your own Ferrari (that you have to be a mechanic for yourself) and it's broken down all the time, but gives you the time of your life vs driving a reliable Honda (Hermes Agent) vs riding the bus (Claude / ChatGPT)

@rayofarsenal @SnazzyLabs a vpn allows you to remote access an apple tv?

I help support my parents back home with tech issues while I'm assigned here in Africa, more than 7,000 miles away. I think this would make for really powerful video content. A lot of people are probably going through the same situation. A VPN has been working well for me so far.

Apple, please create remote access for Apple TV. It is nearly impossible to help my grandparents navigate with their remote (and even worse, type in passwords) from 1,000 miles away.

even opus 4.7 is borderline unusable bro the only reason i use claude is how easy it is to talk to but holy hell is it not thorough. its a fairly dumb model imo. if the 5.0 models do ANYTHING, they should focus more on reasoning than just raw output anthropic is definitely benchmaxxing because gpt-5.4 is better is almost every way. it's just not as pleasing to speak with.

@jesse_vermeulen 5-10 mins? codex runs for hours if you write a detailed spec but when they’re not running, i’m just managing the 9 other sessions or scrolling twitter

honest question: what do people do during the 5-10 min while Claude is running?