VOID

@BLASTPremier @parivisiongg

First CS player to file his taxes mid-tournament #BLASTPremier @parivisiongg

@MicrosoftAI @arena Got rate limited after 5 images but the results are stunning! CC @PurzBeats @ComfyUI

Meet MAI‑Image‑2. Built with creatives, for real creative work. Ranked #5 on @arena’s text‑to‑image leaderboard. Available now: msft.it/6014QUCBe

now available: DLSS-5 anything for free ⬇️ how do you know you hate it if you dont try it first? Sharing the hugging face demo

0/10 to 24/22! Tag someone that needs to learn the true meaning of "never give up!" 👇

@OliverPecek06 @Thorin You can delete this tweet now.

@Thorin Will still make finals though, don't see astralis and aurora being a serious threat

Spirit will probably always be slightly over-rated at a minimum and underwhelm because donk is so good he makes any match look winnable.

@Ozzny_CS2 Run to mid evry round to DIE! Repeat!

donk after Spirit got ELIMINATED from ESL Pro League ‼️

@p4ttiie 6 7

?

@ESLCS @AstralisCS @Team__Spirit Astralis eliminating Spirit was not on my bingo card for 2026! Today donk turned out to be a perfect example of how not to play competitive CS2!

.@AstralisCS confidently defeat @Team__Spirit to secure the final spot in the semis of #ESLProLeague Season 23 13-6 Dust2 13-7 Ancient

i predict gpt-5.5 will release on may 14

@_simonsmith Does it have something to do with "Remote browser data" being turned off. Ignore if irrelevant.

ChatGPT Atlas isn't remembering my GPT-5.4 preference and keeps reverting to GPT-5.3 in the sidebar. Anyone else experiencing this?

WHAT???????????????

@_simonsmith Noticed the same in ChatGPT for Excel. The Charts had numbers interjecting in between. I hope they fix it in the final release though beta version performs well in intial tests.

GPT-5.4 is fantastic at many things, but not design. This includes PowerPoint design in ChatGPT. Like, really bad. And yet... Turns out it's not a model issue. I had Claude Opus 4.6 give it instructions to improve its slide design, and it made something elegant. I asked it to evaluate what happened here: - What went wrong: wrong toolchain (python-pptx + raster chart), no enforced design system, no hard visual QA, technically valid ≠ well-designed. - What fixed it: PptxGenJS + native chart, strict palette/hierarchy/spacing rules, simpler layout, render-to-image QA before delivery. So, OpenAI (maybe @JustinBleuel, maybe someone else?), we can fix this! We just need to give GPT-5.4 a way better PowerPoint creation skill. Since we now have skills in ChatGPT Enterprise, I'm going to try building one.

gpt image 1.5 vs nano banana 2

WHAT???????

me stepping down. bye my beloved qwen.

@PurzBeats Maintaining face consistency and maintaining it when there are multilple subjects in the image is what the AI needs to overcome in the years ahead. It is still not great in all the AI models.

I always use myself as a subject for my generations for a few reasons. 1. I know my face better than anyone's so I can tell if the model is maintaining consistency well. 2. I own my face, I'm not infringing on IP when I use myself as a character. 3. Unchecked narcissism.

@Google @GeminiApp Adding more examples of the torn paper effect from Nano Banana 2

Introducing Nano Banana 2: Our best image generation and editing model yet. 🍌 Pro-level quality, at Flash speed. Rolling out today across @GeminiApp, Search, and our developer and creativity tools.

@Google @GeminiApp Some more great examples from Nano Banana 2.

@Google @GeminiApp Some great examples from Nano Banana 2! Looks Awesome!

Who has a 96GB Video Card, a really solid VPN, and has recently updated their Malware Blocker. Because if that's you: Your mission orders have arrived.