ege

hard thing to get right with voice ai is that there is absolutely no margin for any friction or error, the bar is too high with talking to humans

the best ios engineer i know has no online presence whatsoever he just build the stuff you use thousands of times a day so perfectly and goes on his merry way, bit of a loss for the world tbh

@mecid I wish more apps looked like this: simple, functional, will stay up to date with OS. Only invest in special places in your app to roll your custom UI

Most of SwiftUI apps look same because they use List and its default styling. I've introduced the DesignSystem module in my apps containing custom stuff like headers, divided card, hero card, scrolling surface, etc. swiftwithmajid.com/2022/01/12/mic…

love it when the agent sporadically fails and all you need to do is say "keep going" for it to pick it back up and finish the job, some wisdom in there

@essenciverse The team has cooked that one

@aegeantic I mean the voice version by the way

Been going on a 1hr walk everyday this week at 5pm, speaking to grokkie on voice as I walk (part of job), its crazy how much you can learn so effortlessly these days + you feel like him (palo alto version)

@Baconbrix Bro hahahahah

ok but now combine it with Apple vision models and Expo Agent

Sorry marina folk for blasting Simge on the uhaul truck this afternoon but I was having a moment

@CVincente19 Please flag those we are actively fixing bugs

The feature isn’t perfect as it stops reading at certain points but continues after that text, so I just read the bit it misses then allow it to continue. @aegeantic thanks again for adding this feature.

Being able to listen to X articles is a game changer. It has always been hard for me to read large amounts of text even if I want to, but being able to listen essentially opens up more content I can fully engage with in a reasonable amount of time.

@Designarena The screenshot posted here isn't matching what I see in audioarena.ai/leaderboard, what am I missing?

Introducing Audio Arena Most existing voice benchmarks are approaching saturation - frontier models are scoring 90%+ on nearly every category. Today we've open-sourced a suite of 6 static multi-turn benchmarks designed to stress-test speech-to-speech models on realistic customer-facing scenarios.

@tylerangert @patinasystems Hell yah

@aegeantic @patinasystems Can try that too! Could be a nice cross dissolve or cross blur

Especially in UI engineering, when you break out of the tyranny of “best practices” and aren’t afraid to go low level…you can get magic

@jon 👀

Grok on iPhone

@mil0theminer zirp

Even with an AI model from 2028 this thing still wouldn’t have succeeded

@veggie_eric thx goat

grok-4.20 is live on the xAI API. It's a webdev machine (no purple gradient slop). It also won't break the bank, at $2 in / $ 6 out

@nathantannar4 Liquid glass was a zirp

Here’s to hoping Apple fixes the zoom transition in iOS 27. Did not have the jiggle in iOS 18 when scrolling the scroll view. Is this behaviour expected?

you don’t need more credits you need to —dangerously-text-her more bro

2 most freeing things in life - you don’t have to eat breakfast as soon as you get up - you can eat the same breakfast every single time

@turbo_xo_ @skcd42 👍

Fix pls: Hi, consolidated voice-input UX feedback for Grok (standalone app + Grok-in-X): (1) During voice recording on iOS, the screen can auto-lock; recording continues but it feels broken and can get buggy afterward (eg, Send behaves oddly) so the app should prevent sleep while recording (like ChatGPT) or provide a robust persistent “recording” UI and state handling. (2) After recording, Grok currently transcribes and auto-sends immediately; please instead insert the transcript into the compose box as an editable draft and require an explicit Send, with a user setting to enable/disable auto-send. (3) Support mixed-mode + multi-take composition: the mic should remain available even after any typed text, and multiple recordings should append at the cursor/end of the same draft. (4) Unify transcription behavior across apps: Grok-in-X currently does inferior word-by-word dictation while the Grok app does better full-block audio transcription; please use the full record-then-transcribe block approach everywhere with consistent compose behavior.

I find it surprising that in the era of agent first coding, people still ask for "code pointers" ... just ask your agent lol it would do a much better job than me at explaining the code and overall flow

@maggerbot @paulg Tenderloin pg

Today, Paul Graham himself made a visit to Princeton University. It’s always a rare treat to have a figure like @paulg bestow startup wisdom in our still very traditional, Wall Street-eyed Ivy League bubble. Among the several nuggets of advice Paul shared, one of my most surprising takeaways is that Jessica (@jesslivingston), his wife, doesn’t realize how deeply appreciated her work is with the Social Radars podcast, unearthing so much warmth to the intense and mysterious world of tech startups. I hope my card made it her way across the pond. PG flew in and flew back to England in 24 hours just for us. His words practically raised an entire generation of successful + aspiring founders, so what a gift for a random, gloomy NJ Monday. ♥️