Cosmael / Raphael

Meet Haze! ☁️ She’s my new OC for my hobby drawing. I’ve already got some short comics in the works, so stay tuned for more of her soon!

@SC_Zakalwe @Grummz agree, I think it is a fair score ... and it is a good game, but I wonder why it feels so "pushed" by "influencers", it is definitely not a "perfect 10/10 this is revolutionary!"

@Grummz 30 hours in, so many baffling design choices, still rating it a 6-7/10

I've read so many Crimson Desert reviews and youtube "I don't like this game" reactions that clearly only played the first 3 hours. Very few made it to 7 hours (which lands you in camp if you beeline the main story). All the early negative reviews were BS. They are the true tourist grifters. You can not like the game, it has problems, but this is their job and they didn't do it.

most of the critics I have heard about Outlaws were similar "the game is very boring in its first few hours (empty maps) and combat mechanic is bad" I am not going to defend Outlaws or even say that they are similar, but if a critic is valid for a game, it should be for the other one. If you need hours to start to enjoy it and the combat are bad ... it may be a good a game but not a great one ...

is youtube down?

I am more amazed by Fiat being the most reliable European maker and 4th overall, followed by Dacia. Fiat has always been considered as "terrible quality", if true, this is likely harder for people to acknowledge than Tesla being first.

Finally the #CadillacF1Team reveal!

I agree, maybe my message sounded very negative, but as mentioned that is a very impressive work and yes it did not hallucinate, which I guess was the the first main point. It is more the "accurate summary" or "reliable synthesis" as you mentioned and phrased much better than I did :D

That’s a fair criticism, and I think it’s an important distinction to surface. What you’re describing isn’t hallucination — it’s *coverage error*. The system stays grounded in real papers, but still misses historically foundational work when the retrieval or synthesis objective prioritizes recent, highly-cited, or thematically clustered sources over true chronological provenance. For topics like Light Sheet Microscopy with Adaptive Optics, the *first build-and-use paper* matters structurally, not just as a citation but as the anchor of the field’s evolution. If that anchor is missing, the synthesis can be technically correct yet historically wrong. This highlights a real limitation: retrieval-grounded models still need explicit incentives or constraints for • origin tracing • priority claims • negative space (what must not be omitted) Without that, “reliable synthesis” can still flatten timelines and erase first principles. So yes — this is exactly the kind of feedback that should shape the next iteration: not just fewer hallucinations, but better epistemic ordering of evidence.

Thrilled to share: OpenScholar - our work on scientific deep research agents for reliable literature synthesis -has been accepted to Nature! 🎉 Huge thanks to collaborators across institutions who made this possible!

@pjt_BS damn that's some good progress

☂️🌬️ #TONIN #gamedev #indiedev #indiegame #unity

@morgoth_raven @sama the guardrails are insane, reason why I moved away from it. Even if you have a draft and ask to help make it sounds good, it may try to remove anything that sounds a bit critical ...

@sama "We want feedback" said CEO of the company that stripped warmth from their models and ignored every user request for months. Users have been giving you feedback for months. Remove the router. Ease the guardrails. When will you actually listen?

Tomorrow we’re hosting a town hall for AI builders at OpenAI. We want feedback as we start building a new generation of tools. This is an experiment and a first pass at a new format — we’ll livestream the discussion on YouTube at 4 pm PT. Reply here with questions and we’ll answer as many as we can!

@aisearchio it is funny how it seems it knows but decide that it is not going to do it properly XD

Testing GLM image. It's not that good....

@aisearchio when you order your stars from Temu

@KharayKrayKray mostly on weekends

We need more than 24 hours in a day.

We’re a tiny indie team trying to make our dream game Pizza Kidd with real cheese 🍕 Our Kickstarter is live + we need help spreading the word... A retweet is FREE + only takes 1 second —which helps us more than you know!!! Thank you 🙏❤️‍🔥 — Black Roze kickstarter.com/projects/pizza…

Early Bird Discount is Almost Gone! 🐣 Can’t pledge? Totally okay... Sharing costs $0 and helps more than you’d think... Every RT = +1 hit 💥

✨ Artists! SHOW ME YOUR ART! ✨ • Share your best/favorite artwork below! • Keep it SFW! 👼 • Follow Me if you'd like! • Be respectful and have fun! • Tag your art mutuals, and let's interact 💞 • NO A*I ❌ #artshare #artmoots

I'm confused by his speech. I don't know anybody using AI who says that "art is not important." On the contrary, it's a tool that allows people to create what they have in mind, just by using words. The only counter argument to that is the stealing of data (which is true and not just for "art") and that they could commission someone to do it. However people who can barely afford food have no budget to commission someone for "art", but they can ask a free software on the internet to do it ... that's as simple as that