Stephan

@ZacksJerryRig @BYDCompany @Tesla But why?

Congrats to @BYDCompany for making cars that can literally fly for 6 meters at a time - and overtaking @tesla as the leader in EV manufacturing. Massive accomplishment. 👏👏

@Barflak @StormchaserNL Het blijft maar sneeuwen hier, niet normaal!

@StormchaserNL Stil is! Vrijwel niks in Stadskanaal. 😔

❄️ Lokaal al flink wat #sneeuw! Hier twee plaatjes die ik zojuist heb gemaakt in de heuvels van Groesbeek, bij Nijmegen. De hele nacht zijn sneeuwbuien gevallen: op het gras ligt inmiddels 6-7 cm en daar komt zeker meer bij:

@OfficialLoganK Yes please, would love to try it!

Reply here or DM me :) will add folks in as much as we can

Big upgrade to vibe coding in @GoogleAIStudio lands in Jan, but if you want to test early… 👇🏻

@DJSnM @smartereveryday It never even dawned on me that other people might not be able to differentiate between a hyphen in a domain name and a dot signalling a subdomain. Thanks for this reality check, it's quite educational.

@smartereveryday It's phishing from the domain 'fileappeal-x.com' which is not owned by x if it were legit it would be a subdomain 'fileappeal.x.com' - note the dash vs the dot.

I can't tell if this is a spear-phishing attempt or if someone actually issued a copyright takedown on this image of my birthday cake. Feels like spear-phishing. Does anyone know how to check without typing a password into this box?

@sama This tweet stuck with me and Codex with GPT-5 is getting there for me. Just created a custom tool for analyzing my health data, didn’t touch a line of code. Can easily import a csv and it runs locally for added privacy. Not super pretty, but does exactly what I want it to.

entering the fast fashion era of SaaS very soon

@sama On number 3, l'd say take inspiration from psychology and interviewing. Many good (forensic) interviewers mirror the person they're taking to. I imagine people would like ChatGPT to somewhat mirror their own style. It would probably also help make writing seem less Al generated.

Wanted to provide more updates on the GPT-5 rollout and changes we are making heading into the weekend. 1. We for sure underestimated how much some of the things that people like in GPT-4o matter to them, even if GPT-5 performs better in most ways. 2. Users have very different opinions on the relative strength of GPT-4o vs GPT-5 (just the chat model, not the advanced reasoning one). This is a cool thing you can try: x.com/flowersslop/st… 3. Long-term, this has reinforced that we really need good ways for different users to customize things (we understand that there isn't one model that works for everyone, and we have been investing in steerability research and launched a research preview of different personalities). For a silly example, some users really, really like emojis, and some never want to see one. Some users really want cold logic and some want warmth and a different kind of emotional intelligence. I am confident we can offer way more customization than we do now while still encouraging healthy use. 4. We are going to focus on finishing the GPT-5 rollout and getting things stable (we are now out to 100% of Pro users, and getting close to 100% of all users) and then we are going to focus on some changes to GPT-5 to make it warmer. Really good per-users customization will take longer. 5. The team is doing heroic work to optimize our systems and find more capacity, but still, we are looking at a severe capacity challenge for next week. We are still deciding what we are going to do, but we will be transparent with our principles. Not everyone will like whatever tradeoffs we end up with, obviously, but at least we will explain how we are making decisions. Thanks for your patience with us; we will continue to react and improve quickly!

@codetarded @scaling01 Every wish a business person give to a dev is extremely underspecified. If they really want to replace developers with agents, this is exactly what LLMs need to be able to work with.

@scaling01 so yes, extremely underspecified.

GPT-5 vs Grok-4 exact same prompts, wildly different output in web lmarena

@scaling01 @codetarded Now make it respond to music! I thought it was doing that already and was disappointed that it wasn't.

basically this: Create a stunning, interactive animation of a neural network or brain-like graph structure—use artistic colors, smooth transitions, and beautiful visuals. The page should feel alive, immersive, and impressive, with no buttons—just scrolling or continuous animation. Make it breathtaking. then i just prompted for improvements but Lobster was 100x from the start

@IncCombustion just noticed your 2 thermometer + screen bundle is $1 more expensive than just buying thermometer + screen and a separate thermometer. But it's listed as a $50 discount, which it isn't.

@apples_jimmy @jasonbotterill Here too. Now to think what to do with it...

@jasonbotterill People are slowly getting it now x.com/bughunternerd/…

Wheres my ChatGPT agent sam? People are inventing new science while im still waiting for the rollout. Will I also be the last of the GPT-5 rollout sam?

@apples_jimmy @tetsuoai @OpenAI MONDAY, MONDAY, MONDAY!!

@tetsuoai @OpenAI “ In the upcoming Mondays “

Now rolled out to 100% of Pro users. Due to higher than expected demand, Plus and Team users will begin getting access Monday.

@Angaisb_ And running locally on modern hardware without too many compromises. I want my phone to be a real-time guide in a museum or translator on a trip abroad.

Current models are really capable, what if instead of getting more intelligence we focus on making them fast? Imagine Sonnet 4 making websites in the blink of an eye

@MichaelSpo50061 @apples_jimmy Could very well be that other models run into context limits and just fail hard at some point in time, while Grok lasts longer but gets progressively worse over time/context. Can't find details on this though, so it's just my best guess.

@apples_jimmy The seams misleading looking how the other ones are still going straight up except for Claude leveling out so why wouldn't they continue the time with the other models? I think I need a lot more context or it's just not a fair comparison.

Grok 4: Still no wall. 50.7% with Grok 4 heavy on humanity’s last exam 41% with tools 26.9% without tools. " Grok 4 potentially better than phd level in every subject no exceptions " " discover new technologies maybe this year and new physics certainly within 2 years "

@apples_jimmy @bobmcgrewai Sinus is more interesting than a straight line though.

@bobmcgrewai Very annoying to see in action.

This but for everything

@hi__hassan @sama @ladoger @yacineMTB

@sama @ladoger @yacineMTB you've my blessings sam

i think computers should be cute

@AndrewBenson @levie Nor will it have access to things like intuition or information that isn't documented but passed down. For example when working on a car, there's always a critical piece missing in guides that I have to figure out myself. Some of this info is just passed on from senior to junior.

@levie But it won’t have access to private data, which remains a moat

@apples_jimmy @pranaveight Thanks for your patience, Jimmy.

@pranaveight Haven’t had that before

Canvas sharing is here! I one-shotted an app to plan my gardening for this year chatgpt.com/canvas/shared/…

@sama Oh, and also, I want GPT with me on standby. Like when driving in a foreign country to tell me what a sign says, without having to pull out my phone. Or to tell me about a POI I'm looking at. Or to fact check. So I guess I want it inside smart glasses or something.

what would you like openai to build/fix in 2025?

@platformpilot89 @apples_jimmy @sama @ti_morse

@apples_jimmy @sama @ti_morse So much winning, you're going to be tired of winning. You'll say "please Mr President, that's too much winning! We've had enough winning". But we're going to keep on winning & winning!

@sama I really like the idea of using advanced voice as a translator, since both TTS and translation are great. Unfortunately it's too sensitive for that, since any noise in the background will stop it from speaking and the translation will be cut off. Translator mode would be great.