alocom

@signulll But Sam is Jewish!

@petergostev Windows!

Added the limit rings to GitHub, this is independent from the pet, so you can use whatever pet you like: github.com/petergpt/codex…

Got my dog as a Codex pet, but more interestingly got Codex to add the rings to show my Codex limits. Outer ring is 5 hours, inner the weekly one

@arrakis_ai Microsoft edge!

Use Codex directly inside Chrome. Chormex just dropped v0.1.3 and it’s kind of wild. You can now connect Skills, plug into MCP and plugins, generate comics from the page you’re on, highlight any context and ask questions instantly, get site-specific suggestions, use Plan Mode, even run /goal workflows. And the best part? It syncs with your ChatGPT account. No friction. If you’re already using Chrome, there’s basically no reason not to try this. Also… it’s 100% open source.

GP-5.5 on SimpleBench @kimmonismus @haider1 @AiBattle_ @scaling01 @koltregaskes @MatthewBerman @chrisgpt

@kierajmumick @syedmuzayan @OpenAI Is it coming to Windows?

@syedmuzayan @OpenAI Right now, yes!

Excited to announce the first major product I’ve worked on at @OpenAI : Computer Use in Codex! Our team has been hard at work bringing this magical experience to you over the last couple of weeks Let us know how you use Computer Use! Some of my favorite details 👇

@JamesZmSun @AndrewMayne

@AndrewMayne We haven't plugged in the browser-use agent yet, so the only agent workflow that Codex knows for the browser is iterating based on your comments! We'll ship an update next week that lets Codex fully drive the browser.

We are super excited to launch the in-app browser inside Codex with comment mode! View any web pages & iterate with your agent quickly with just point and click. Codex will automatically capture a screenshot, the DOM element, and feed it as precise context to your next chat. No more switching between browsers, dragging screenshots, and wrangling with underspecified prompts. It's great for front-end development of apps/pages, but also very useful if you have documentation pulled up on the side and just want to ask a question!

@scaling01 Come on, don’t be mean!

every week he apologizes for his latest fuck up or lie imagine in a year: "sorry guys, we didn't mean to cause $100B in damages through cyber attacks. It has been a tough day for me, so it's not really my fault and you can't blame me"

@TadawulFeed إدارة سيئة أهلكت الشركة وأفلستها! خسارة 95٪ من حقوق المساهمين جناية تستحق التحقيق

اعلان الشركة السعودية للطباعة والتغليف عن النتائج المالية السنوية المنتهية في 2025-12-31 saudiexchange.sa/wps/portal/sau…

@pashmerepat Use Codex to create office work—Word docs, PowerPoint presentations, Excel spreadsheets, and PDF files—in Arabic. I’ve already updated custom skill files for that, to support Arabic. @thsottiaux @reach_vb @romainhuet @heccbrent @ajambrosino @dkundel

Everyone talks about Codex for coding. I want to hear about the other stuff. If you're using it beyond writing code, what's your workflow?

@ShakeelHashim @AndrewGinns @TheRealAdamG If Anthropic and Dario win the contract, can they stop anything from happening?

@AndrewGinns @TheRealAdamG But those engineers can’t actually do anything. They can’t stop anything from happening.

What we know about the OpenAI-DoW deal: OpenAI agreed to the terms Anthropic rejected. The terms include an "all lawful use" clause. The contract "references certain existing legal authorities" which the govt claims prove that domestic mass surveillance is already illegal. There are "certain mutually agreed upon safety mechanisms" — we don't know the precise details of these, but it looks like that includes a "monitoring harness" and Forward Deployed Engineers, which, per one OAI employee, "seem to offer broad powers for openai to interpret the law." Perhaps! But ultimately OpenAI isn't in charge of how this model is used, and it has signed a contract that allows the Pentagon to use the model for all lawful uses. The Pentagon may claim that domestic mass surveillance is illegal, but that is on a very narrow definition of domestic mass surveillance — the Snowden files showed that the Pentagon is very willing to play word games to pretend that domestic mass surveillance isn't what it is. I think the default assumption should be that OpenAI's "safety mechanisms" won't actually do anything, because ultimately they've agreed that their models can be used for anything lawful. Which is, in effect, the same thing as letting the Pentagon use their models for domestic mass surveillance.

@LAMMMAH ما أفضل صكوك؟

🚨 تنبيه .. من يرغب في شراء الصكوك ( أي صكوك ) فليتأكد من عمولتها من الوسيط الذي يتعامل معه .. فالوسطاء يتفاوتون .. وليست مثل الأسهم محددة .. أمس تواصلت مع ثلاثة وسطاء : واحد طلعت معه العمولة 1% .. والثاني 0.255% .. والثالث قال ماعندنا تداول صكوك أصلاً .. وأخشى أن يكون وسيطك يأخذ نسبة عالية تذهب بالعائد .. كيف تعرف كم العمولة ؟ كلّم الوسيط واسأله مباشرة .. أو جرب بمبلغ 1000 ريال وشف كم يأخذون قبل لا تشتري بالاحتياطي الفيدرالي .. #سوق_الأسهم_السعودية

@NanoBanana أنشئ صورة واقعية جداً، لشارع طويل ممتد في مدينة مستقبلية، وفيها الكثير من الأضواء، وفيها لوحة إعلانية مكتوب فيها: مرحبا بك في المستقبل

@javiruli22 @kimmonismus yap

@kimmonismus Am I the only one who feels like GPT-5 is dumber than Claude / Grok?

This is crazy: GPT-5 medium outperforms o3-pro high at ~90% lower cost. Intelligence too cheap to meter. I can only repeat it again and again.

@dkundel How can I use the OpenAI API key in Codex vscode without using a ChatGPT subscription?

We’ve seen some confusion around signing into Codex with ChatGPT Edu or Enterprise accounts. We are currently working on an issue that’s blocking the sign in. Once resolved, as long as your workspace admin enables signing in with ChatGPT, you should be good to go. Apologies for the delay here.

@scaling01 But you're coming for OpenAI, while Anthropic is jacking up prices and cutting the limits, and you say nothing about them?! Your claim is looking real shady.

I believe AI is the most important technology in history this is why I have insanely high standards for leading AI labs if they act greedy, rush products launches or present something that they didn't think through, I'm going to call out their bullshit and raise hell if I have to

@btibor91 GPT-5-mini GPT-5-nano ??

Retested how much reasoning "juice" GPT-5 uses in the API vs ChatGPT, depending on the action you take Tested using API, ChatGPT Plus, and ChatGPT Pro as of today

@lehai0609 @btibor91 GPT-4.1 GPT-4.1-mini GPT-4.1-nano 1m context windows

@btibor91 They avoid the questions about context windows. Seems like that's the issue they CAN"T fix. Personally I can't accept that. It's time to move my money to Gemini & Anthropic.

Summary of the GPT-5 AMA on Reddit with OpenAI's Sam Altman and some members of the GPT-5 team (August 8, 2025) Launch problems and technical fixes - GPT-5 launched with significant technical issues including the autoswitcher being down for a chunk of the day which made the model seem way dumber, but they're now making interventions to the decision boundary so users get the right model more often and adding transparency about which specific model is answering each query - The massive scale rollout caused API traffic to double in 24 hours and created serving hiccups for hundreds of millions of users, but the team expects everything to stabilize within a day or two while they continue working on fixes and are doubling rate limits for Plus users as they finish the rollout - They had bugs with the model router not properly directing coding queries to the thinking model, which caused GPT-5 to seem like a downgrade for programming tasks until users manually triggered reasoning mode or asked the model to "think hard" - The livestream presentation had human error with misleading bar charts due to people working late and being tired, though the actual numbers in the blog post and system card were accurate - The model switcher currently decides based on domain and complexity but doesn't search the web first because that would increase latency too much User access and model options - OpenAI heard loud feedback about bringing back GPT-4o and will restore access for Plus users while monitoring usage to determine long-term support - They're working on GPT-5-mini to restore the same overall reasoning usage limits that users previously had with o3, o4-mini-high, and o4-mini, since removing mini significantly decreased total reasoning capacity, and GPT-5-mini is being gradually rolled out to users - Plus users are supposed to have unlimited access to GPT-5, and anyone getting rate-limited is experiencing what they consider a bug that needs investigation - The team is exploring whether users need access to both 4o and 4.1 or if just 4o would be sufficient, and they're considering a pricing tier between the current $20 Plus and $200 Pro options - They're making UI changes to make it easier to manually trigger thinking and considering adding the ability to force thinking through custom instructions Safety and filtering improvements - They're actively fixing over-flagging issues in biological safety that were incorrectly blocking legitimate academic research about gene therapy and bioengineering, while maintaining necessary safeguards against dual-use risks through their Life Science Research Special Access Program for vetted users - The content filtering system that was flagging users for learning about historical topics like artists or historical figures is being improved, though they acknowledge the challenge of distinguishing between helpful education and harmful content Coding capabilities and tools - Codex CLI now supports GPT-5 via paid ChatGPT plans, with usage limits designed so Pro users rarely hit restrictions while Plus and Team users can handle several 1-2 hour coding sessions weekly, and rate limits reset every 5 hours with weekly totals - They're planning to integrate GitHub directly into normal ChatGPT conversations without requiring Deep Research or Agent mode - Canvas for coding is getting continued improvements with significant upgrades to Codex coming soon, and GPT-5 thinking is particularly strong at complex reasoning problems, long refactors, and building complete applications from zero including functional frontends - Tool use and function calling in the API has been improved for GPT-5, they're considering GPT-5 Pro for the API but note it would be slow and expensive, and there are no concrete plans to bring GPT-5 Pro to the Codex CLI - Internal developers use various coding tools including Codex, Cursor, and are building internal debugging tools for training runs, which has gotten much easier with GPT-5 Model features and future plans - GPT-5 represents huge improvements over GPT-4 in reasoning, creative writing, instruction following, reduced hallucinations, healthcare, and visual perception, with GPT-5 thinking expected to have better and more interesting writing quality than GPT-4.5, and the team believing the GPT-4 to GPT-5 jump is as significant as GPT-3 to GPT-4 - The model has quirky characteristics including a strong preference for purple in website designs and a tendency to say "let's craft" in its chain of thought, which became internal team rallying cries - Memory improvements are coming where the model will remember steering preferences once set, and they have exciting memory-specific developments coming soon - They launched a new voice model that's better at instruction following and responsiveness, with advanced voice mode coming to all free users and plans to make it more multimodal with useful information displayed - Context window improvements are desired (they wanted 1M tokens for GPT-5) but are constrained by compute costs and GPU capacity, though Pro users get 128k context limits while Plus users get 32k - They're working on unifying reasoning and non-reasoning models into one single model in the near future, and considering token-based pricing rather than message limits - Future plans include screenless access through voice and possibly neural interfaces much further away, with the goal of helping users spend less time on devices while getting more value - The team tested different model snapshots including Summit and Zenith on various platforms, with Summit ultimately outperforming Zenith by wide margins on leaderboards and internal testing

@chetaslua @kimmonismus @btibor91 @adonis_singh @chrisgpt @BennettBuhner @aidan_mclau @DeryaTR_ @nadzi_mouad @wesroth @flavioAd for io devices!!

So what is Aura? “Aura” is the codename for the lightweight, voice-first/minimal-Browser version of ChatGPT that powers: the “Voice Mode” experience on mobile, the embedded ChatGPT that pops up as a small overlay (“mini”) in partner app. Removes clutter – the sidebar, most header actions, and some menu items are suppressed. Optimises for voice & small surfaces – larger touch targets, a floating microphone button, and subtle theming tweaks.

@HAlmnhli يفتح معي وأدخل البيانات بعد الضغط على زر تدقيق، تظهر رسالة: الخدمة غير متوفرة

التسجيل شغال فقط عن طريق الرابط ولابد تكون مسجيل دخول من قبل وفاتح على نور الجديد noor.moe.gov.sa/Noor/Admission… #نظام_نور #تسجيل_رياض_الاطفال #تسجيل_الروضة