Neo

@thdxr @kokialgo @maria_rcks @uzairakrum @karthiknish @ashirwadsingh_ @rhacius @Ananth7e @nigadamus @diegodamval @0XB34RDYB0Y @Dan_VAlves @TanmayGoel442 @InnoboSJ @Burgess_33 @_kittylove3 @alexlcpereira @NeoReplicante @Ableezz @Jint3x @HareeshXAi @defyus @Nateemerson Thanks! e-mail sent :)

@kokialgo @maria_rcks @neocph @uzairakrum @karthiknish @ashirwadsingh_ @rhacius @Ananth7e @nigadamus @diegodamval @0XB34RDYB0Y @Dan_VAlves @TanmayGoel442 @InnoboSJ @Burgess_33 @_kittylove3 @alexlcpereira @NeoReplicante @Ableezz @Jint3x @HareeshXAi @defyus @Nateemerson @andreafspeziale you guys caught ryan doing the macarena if you want your free month of OpenCode Go send an email to dax@anoma.ly subject line "Macarena" from the email you want access with

@weswinder @thsottiaux Same for me.

i still haven’t received a single banked codex reset am i the only one who didn’t get any? @thsottiaux what’s going on here?

@theo so they have the implementation then? they could use it 👀

This page is from April and has nothing to do with the Fable ban. Misinformation.

@DanielSmidstrup i'm sorry Dan i'm afraid i can't do that

prove me you're not an ai

@markproduct i'm sorry Mark i'm afraid i can't do that

prove me you're not an ai

@mark_k @sama Maybe the trial has something to do with it? A lot of things came out during the trial, which changed the whole story.

Looks like @sama Altman wasn't too happy with the biopic about him... Amazon just canceled distribution of the already finished movie. 🍿 We might still get a chance to see it, if another distributor can be found.

@robinebers it's really good for code review (5.1 was too). I run that, and have 5.5 judge which bugs are real and which to actually fix. GLM produces some false positives, but finds a lot that the others don't. So you need to also evaluate the report it gives.

OK i'll put GLM 5.2 to the test will report

@thdxr The macarena of course

first 25 people to guess what he's doing here get free OpenCode Go for a month

@0xRemakw @thsottiaux i think they broke caching, not reduced usage.

CODEX LIMITS REDUCED 50% can someone verify if this is true ? @thsottiaux

I want a way to turn off you collecting data for training, without loosing persistent threads throughout Gemini apps. Until then, no matter what models you have, they’re not valid for any real work, and honestly I wouldn’t use it for personal use either. Even the API will collect some data.. I want to be able to turn that off.

@clanker_ no plan on giving up, but feedback welcomed as you try out everything we shipped at IO : )

Google should give up now, I/O was embarrassing

I want a way to turn off you collecting data for training, without loosing persistent threads throughout Gemini apps. Until then, no matter what models you have, they’re not valid for any real work, and honestly I wouldn’t use it for personal use either. Even the API will collect some data.. I want to be able to turn that off.

@OfficialLoganK @theo Since feedback is welcome, can you make it so we can turn off that you train on our data, but we still keep our threads in general in the Gemini apps? It makes any model improvements you make useless for a lot of stuff. I can’t use Gemini for anything serious until you do.

@theo feedback always welcome : ) we will keep shipping and iterating!

Just filmed a Google crashout video. One of the best I've done in a bit. Hopefully they won't demonetize my YouTube channel like they did last time 🙃

@iamdavidhill @opencode So when are you releasing the update with gpt 5.7 pro? 🤔

we’ve been busy redesigning the desktop app from the ground up the beta taught us a lot most importantly, that left sidebars suck hello tabs 🔊

@sama What if it gets more? No reset then? 😬

if this tweet gets 1 like, tibo will reset codex rate limits

@thsottiaux I also have issues where, as a session becomes longer than 2-3 compactions, it just won’t send anymore, and slows down so much, that it’s unusable. Did your caching break or something? I get retries, slow connection and websocket errors. I can’t do any task longer than 10-20 min.

Seeing issues where usage limits are out of sync for some Codex users. Apologies and team is investigating.

@robinebers You should have plan ready for a mid week reset at some point 😂

i’ve done it burnt the equivalent of 10x $200 Pro learned so much this week but damn this was exhausting 36.8 billion tokens much harder than I thought

@VilasMagare5 @thsottiaux Yeah you’re right. Missed that it could do a heartbeat automation within a thread.

@neocph @thsottiaux loop already works. it just don't have the /loop command

The master becomes the mentee. At last, Claude is now copying Codex. But you cannot out-accelerate GPT-5.5.

@digitalogaa @thsottiaux You’re right, it does! I missed that one. You can set up a heartbeat automation within a thread. Thanks!

@neocph @thsottiaux Codex already knows how to loop. It doesn’t have to be a feature. Tell it to loop and it would. It’s smart enough

@digitalogaa @thsottiaux Loop is heartbeat system in Claude code, it isn’t in codex. Or did I miss it? I’ve used loop in Claude code via remote control to check in hourly on codex in the past, and keep it going.

I understand. I’ve done a LOT of rendering and 3D work myself, at massive scale. We’re in an in-between stage right now, where it’s easy to get something out, but not quite exactly what you want. And we can miss out on the fine control and satisfaction that 4 days of perfecting a set of materials you’re gonna reuse constantly will give, and the result it’s going to give. I found that for some things I still need to do it the old fashioned way. For other things, which I just don’t want to spent hours or days on, I love the much faster tool that AI can be. And you can absolutely still iterate with images, it just takes time and practice to be able to explain in words what you have in your head. I think you got a pretty good result. Only you care that I didn’t take you 5 or 15 hours, users won’t 😊 And there’s the fact, you can keep going down that path. Some people missed the satisfaction that painting and drawing everything by hand, when the age of 3D and CAD came.

@rahulbhadoriiya well you did describe a design process - just a different tool for producing the final result... The design didn't happen in the AI, the design happened before that and in between each attempt.

Question to my fellow designers: This is an app icon I made using AI. The process was actually very similar to how I normally design. I collected multiple references, explored directions, and built the visual. The only difference? Instead of using Blender, I used a node-based AI tool. Now here’s my question: Do you think this is good or bad? Personally, I like the result, but it didn’t feel satisfying. I’ve spent hours in Photoshop extending backgrounds or tweaking tiny details. This one, I could’ve also made in Blender in maybe 2–3 hours. So now I’m wondering: Is this still design?