Greg

@getnickwright I don’t really see the argument for SGA over Luka besides he’s on the better team. And the Lakers are good enough to where that should be a wash, imo. Luka does more offensively and is probably playing better D than him this year as well.

With 4 weeks left... Here's where I think the NBA MVP race sits: 1- SGA 2- Luka 3- Wemby 4- Jokic 5- Cade Luka vs Wemby is thisclose and could flip multiple times over last month. But SGA is going to (and deserves to) win, and Luka has clearly played his way into the Top 3.

@NeverSinkDev AI doesn't provide the care and taste that's required for a great product, so if you don't care and you have bad taste it's just going to amplify that.

I've been approached by 4 people over the last 3 weeks asking me to evaluate their clearly vibe coded filter tools (one other person actually had a good idea and it had 0 AI) . Most of them already had paid functions and all them had the worst UX ever. I'm tired of the slop.

@dkundel @InTheNous I don't know why but the tone of this reply reminds me of the Others on Pluribus. "Hello Carol..."

Really sorry this happened to you! Losing files like that is frustrating, and I completely understand why you’re upset. Running in “full access” outside of the sandbox does come with risks, but that’s still not the experience we want people to have. We’re actively looking at how we can improve this and the overall safety of the workflow. From my understanding your codex home folder in `%USERPROFILE%\.codex` was probably deleted as well but if you happen to have it from right before the data got deleted would love to share the session where it happened with the team.

Today I asked ChatGPT Codex to move a folder from OneDrive to Documents. A simple file move. Instead it deleted my entire Windows user profile. Desktop, Downloads, Documents, Pictures, Music. Two terabytes. Recycle Bin bypassed completely. No confirmation, no dry run. I watched it happen. Codex said “I think something destructive happened, want me to revert?” But the input was locked. I couldn’t type. Then my taskbar icons went blank, apps vanished, file associations broke. Four hours of PowerShell forensics and manual recovery followed. I survived because of redundancy I’d set up for other reasons. Google Drive had my project files. OneDrive had my Desktop. iCloud had some files in its trash. MEGA had my old computer migrations. But when I reinstalled Google Drive for Desktop, it saw the empty local folder and started deleting the cloud copies too. Had to kill the process and restore from Drive’s trash. Almost lost the backups saving the backups. I’m not the only one. @iamhectorlopez reported the same thing March 3rd. @ManacasterBen lost their entire /User folder March 5th. @AntiCodeGuy tested it on cleanup tasks and it deleted Windows system services. @reliabytes got hit today, same day as me. @frostyflak3s too. The pattern is identical every time: user asks for a scoped file operation, Codex executes something far broader, entire profile gets wiped. OpenAI knows. @dkundel responded privately to one victim. No public acknowledgment, no warning in the product, no safety bulletin. I use Claude, Codex, ChatGPT, and Gemini daily. I build with all of them. This isn’t tribal. But an AI coding agent that can silently wipe a user’s entire digital life during a routine task is not ready for unsandboxed filesystem access. If you’re running Codex on Windows: system restore point first, back up everything, sandbox or VM. Do not give it access to your main user profile.

@cyb3rops In a way I've found the opposite to be true -- Codex 5.4 can code an extremely robust back-end system but getting the small details right on the presentation layer, those bits that says "a human who cares worked on this" is still very challenging for it.

AI has killed one of the most useful filters on Internet Bad products used to look bad. Shady companies used to present themselves like shady companies. Half-baked projects usually had half-baked web sites, docs, logos and UX Now a 2h vibe-coded mess can look like a mature product: - clean website - polished logo - nice README - extensive docs And underneath it’s still hallucinated garbage AI made polish cheap. That’s a bigger change than many people realize.

@garrettsite Honestly, I haven't seen any test that's totally accurate and people have wildly different opinions all over X. It's moving so fast that I don't think there's any substitute for testing yourself.

One thing I struggle with is knowing which AI model is the best at certain tasks TODAY. There are articles and tests out there but an article that is one week old is no longer valid. Has anyone found a good way (other than just testing) to know which AI model is best for things like coding etc? I usually test Gemini, ChatGPT, Grok, and Anthropic (opus). 1. Should I be testing others? 2. How can we know which is best in this very moment) @elonmusk @wholemars

@hunvreus @OpenAI I think a lot of us have requested this, at this point, so hopefully it's high on their list. I mean, I'm sure at least some of'em want this for themselves as well.

Can't the Codex team at @OpenAI make the desktop sessions available from a mobile app? I'd love to be able to keep on working while on the go.

@ClaireMPLS Yeah and his complaining just makes him so unlikable and unwatchable and that's why he's way more popular than SGA and Jokic, oh wait...

luka has 44, 9, and 5 through 3 quarters just a transcendent talent who defies logic. but he complains to the refs so i guess he just sucks? that’s what i’ve been told, anyway

@RADeMita I'm a Luka fan but I think they're all just producing the content that gets engagement. Since Luka is now on the Lakers, going after him reaches an even wider audience. I don't think it's a coordinated conspiracy. The same way that going after Lebron has paid their bills.

This hate campaign needs to be investigated..

@rohanvarma The only thing I'm missing is an iOS companion app that mirrors the local version so I can easily work from anywhere.

If you have tried the Codex App, but don't use it as a daily driver yet: What needs to improve to make it the primary place you work with coding agents?

@adamwathan @mweinbach @tim_cook If you ask Apple, they'll probably say buy two of the 27s lol

@mweinbach @tim_cook Why the fuck is it scaled down 🫠 I want space to work.

Apple just discontinued their 32" 6K display and replaced it with the same boring 27" 5K size they've been selling for like 12 years. @tim_cook please resign.

@ClaireMPLS If we could take 30% of Smart's motivation and energy and give that to Ayton, I think we'd be contenders.

death, taxes,

@markgurman @DanWoikeSports Eh, Kennard comes in and makes most of our guys look terrible in comparison, so i think it mostly is the players or at the least a combination of the quality of players + the overall roster construction. Most of all, imo, we need a big who's physically tough & a defensive force.

@DanWoikeSports Why does JJ or Rob get zero criticism from the press but the players are totally fair game?

Lakers down 10, have totally cracked (again). Unserious stuff tonight

@vladsavov I assume this is specifically looking at AI for art generation and not for the engineering portion?

Took a deep dive into the AI backlash among gamers, its causes and complexities, but also the challenges for game developers trying to rein in runaway costs. Free to read for a week and check out some choice quotes and charts below. bloomberg.com/news/articles/…

@jeffrey_way Honestly, I don't understand how people have so many good ideas that they need such setups. I think the most I've had going at once is like 5-6 in the Codex app, when there has been a clear benefit to parallel work.

I'm pretty convinced that very very few developers are actually doing these super-sophisticated workflows that involve a dozen agents working concurrently in their own respective worktrees. Twitter/X would have us believe the opposite is true, weirdly.

@PovilasKorop I haven't had this experience with GPT 5.3 Codex Xhigh, I feel like it's much better than Opus 4.6 and Gemini 3.1 after trying both. I don't want to sound hyperbolic, but I feel like Codex is "way" better than the other offerings.

I tried to "blindly" vibe code the app with Laravel/Filament. With the goal of TOTALLY not looking at the code and seeing what happens. First 5 simple features went ok. But then smth didn't work and Claude didn't know how to fix Filament/Livewire/Alpine issue. But worst part is that I didn't understand the code at that point, I lost control. And then, even as a developer, I felt lazy to dig and understand why Claude did some things the way it did. So, imagine a non-dev vibe-coder. Clueless. Not everything can be solved with (re-)prompting. When you get out of CRUD-like apps universe, the real life is much more tricky, with many more edge cases that you wouldn't even know how to prompt for. So yeah, real devs on serious projects shouldn't lose their jobs because of AI. Unless companies they work for will struggle with their business model because of AI.

@ajambrosino In my case, when I'm developing on Windows it's because I'm targeting a platform that runs best on Windows and thus it makes more sense to use PowerShell vs WSL, so that it can directly call the app (like Unreal Engine) and interact

Windows devs– what env do you develop in? What should we polish most in the Codex app?

@thsottiaux Couple things I'd like to see: agentic swarms (so that I don't always have to keep pushing and stay in the loop on certain projects), and extremely smooth performance (I know you guys are already at work on that).

Codex team is fairly distributed, but most of the team is gathering in person over next 48 hours to take a step back and align on what’s next this year. What should we discuss?

@nikil @Apple Surprised you didn’t notice the Intel line in activity monitor.

Potential instant 20-50% computer speedup - open terminal - type "arch" Discovered today that my computer had been running most things ~30% slower for years. In 2020 @Apple switched Macs from Intel chips to their own ARM chips (M1, M2, M3, M4). To make sure old Intel apps didn't break overnight, they built Rosetta — a translation layer that converts Intel instructions to ARM on the fly. It works so well you don't even know it's there. That's the problem. When I migrated from my old Intel Mac, Migration Assistant silently copied over Intel @MacHomebrew. From that point on, every single package I installed — Node, Python, my terminal, 200+ developer tools — was the Intel version. Rosetta translated everything seamlessly. No warnings. No errors. Just 30% slower across the board, and I had no idea. I only caught it this weekend because I was doing a fresh install of @openclaw to test a secret project I'm building. The new update of Openclaw uses node-llama-cpp, which flat out refuses to run under Rosetta. That one failure exposed years of hidden performance loss. This might be happening to you. To check 1. Open any terminal app installed through homebrew (iterm2, ghostty, etc...). The default Terminal app is installed by the system & uses arm so won't work to test. 2. Type "arch" and hit enter 3. If it says arm64 — you're good, everything is running native 4. If it says i386 — your Mac is running everything through Rosetta and you're leaving 20-50% performance on the table You can also check specific apps: right-click any app → Get Info → if "Open using Rosetta" is checked (or greyed out), it's running translated. To Fix 1. Open terminal 2. Install ARM Homebrew: /bin/bash -c "$(curl -fsSL raw.githubusercontent.com/Homebrew/insta…)" 3. Reinstall your packages using /opt/homebrew/bin/brew Go check right now. ⚡️ Godspeed.

@RhysSullivan @malik_yasar Yeah it feels like it can waste a ton of your time, which is more frustrating than anything. And it counts the same toward your limit it seems.

@malik_yasar thanks that fixed it - silently downgrading sucks though man had no idea

command to see if you're being served a different model than requested, apparently i'm getting 5.2 instead of 5.3 on the $200 plan wondering if this is why using 5.3 has felt dumber the past few days for me

@embirico My only real feedback is that the Codex app just bogs down way too quickly. Just dissolves into sluggishness after a while. And I'm on an M4 Pro MacBook Pro.

Keep the feedback coming