Sword09
764 posts


Hotfix Notice: 2026-05-01 @10:00 (UTC+8)
Hey Appraiser,
To improve your gameplay experience, we'll be rolling out a hotfix on 2026-05-01 from 10:00 to 14:00 (UTC+8). You may experience a brief lag, crash, or disconnect during the update. Once it's complete, just log back in to finish updating. Sorry for the inconvenience!
Update Details
◆ Fix an issue where "Social Settings" in the Profile might encounter errors.
◆ Add missing English, Japanese, and Korean voiceovers for Aurelia on Android.
◆ Optimize localization for several languages.
◆ Optimize the gameplay experience for Pink Paws Heist and improve the production efficiency of Paw-Paw Coins.
◆ Fix data errors for certain furniture items.
◆ Fix an issue where Jiuyuan could not trigger Basic Attack: Info-Guided Arts in certain scenes.
◆ Fix a potential NPC spawning error during the task "Go to the Bureau of Anomaly Control" in Episodes: A Research in Scarlet.
◆ Fix an issue where the quest marker might disappear during the quest "Dolores bestowed life upon the Puppets" in Episodes: A Research in Scarlet.
◆ Fix an issue where the Spinoff Quest "Tomato Jelly Rampage: Live Report!" could not be accepted even when requirements were met.
◆ Fix an issue where the Ultimate could not be cast during the quest "Track the Tomato Devil" in the Spinoff Quest "Tomato Jelly Rampage: Live Report!".
◆ Fix a progression blocker in the quest "As Expected?" within the Spinoff Quest "A Legendary(?) Band's First Page".
◆ Fix an issue where subsequent quests failed to trigger after completing the tutorial quest "Sales Pitch or Scam?"
◆ Fix occasional game crashes occurring under specific conditions.
◆ Fix a UI overlapping issue on the Arc interface.
◆ Fix a selection error in the Message interface when using a controller.
◆ Adjust the photo mode button mapping for controllers.
◆ Fix an issue where prices incorrectly displayed in USD when opening the Mall for the first time on PC.
◆ Fix an issue with camera jitter when driving down steep slopes on PlayStation 5.
◆ Fix other known issues.
NTE is officially launched! Download now >> nte.onelink.me/9g76/l78pwij6
#NTE #NevernesstoEverness #NTELaunch

English

@Sword09D @AniTVOfficial Nah my third world shit hole doesn't care, I torrent without VPN and I've never had an issue.
English

@Sword09D @AniTVOfficial You know what else brings down piracy sites, publishing them on social media.
English

@Sword09D @AniTVOfficial assert your dominance and torrent infront of your isp
English

@Ryuseikai_ @AniTVOfficial If it bothers you to switch websites every few months / years, you should download media from torrent or usenet since they usually don't go away. Also, archive from physical media (blurays,disks..) you own or from the internet
English

@Ryuseikai_ @AniTVOfficial 800 views are not going to change anything, Reddit and other websites are full of recommendations like this. It's just a false sense of control. Piracy websites will go down (and maybe back up then down again) that's their destiny
English

@Sword09D @AniTVOfficial id rather raw dog it then use proton
English

@AniTVOfficial For manga/manhwa etc I like mangafire.to, not sure if there are specialized good torrent indexers for that
English

@AniTVOfficial You know what won't turn into a corporation ? Torrents !
1. Use a VPN (Proton for example)
2. Go to torrent indexers such as nyaa.si for anime
3. Download torrent using qbittorrent
4. Profit
English

@Nintendeal I like the form factor of the Switch 2. Switch 1 always felt too small in my hands and I never really enjoyed the experience sadly.
For a truly portable handheld though, it's the Ayn Thor / DS form factor which wins 100%
English

> It is 2030.
> Wake up and grab my phone.
> PLEASE SCAN YOUR FACE TO SEE THE WEATHER
> Finish the scan, sigh, walk downstairs. Try to open the smart fridge.
> PLEASE SCAN YOUR RETINAS
> Oblige, wait five minutes for the AI subsystem to determine my age.
> YOU ARE OVER YOUR SUGAR QUOTA FOR THE MONTH. AVAILABLE MEALS: BEANS.
> Swear under my breath. This is the fifth time this week my only option was beans.
> Open the fridge.
> DING DING DING! HAVE YOU CONSIDERED GOING TO TACO BELL FOR A NEW -
> Slam the fridge shut. I forgot to pay to remove ads from the fridge last month.
> Sit down at the table, eat my cold beans (the microwave fingerprint scanner is broken).
> Get into my car to go to work.
> PLEASE SCAN YOUR HAND ON THE CENTER CONSOLE TO BEGIN DRIVING.
> Place my hand down and roll my eyes.
> HERE IS YOUR LIST OF APPROVED DESTINATIONS.
> Every destination is Taco Bell.
English

@TheIshikawaRin Because they talk out of their ass they have no idea what they're talking about
English

@ray5ar @KimNoel399 C'est sûr, mais sans commuication publique on ne peut pas savoir :)
Français

@Sword09D @KimNoel399 Je pense que c'est plus que ça. l'implémentation joue aussi. sur llama.cpp et sur la création de gguf il y a souvent eu des bug qui rendaient les modèles debiles.
Français

Cette entreprise canadienne prend le parti d'imprimer directement dans la puce les poids pytorch des LLM. Résultats, une inference de dingue à 16 000 tokens par seconde !


Taalas Inc.@taalas_inc
24 dedicated people. $30M spent on development. Extreme specialization, speed, and power efficiency. Today we launch Taalas’ first product. Check it out: Details: taalas.com/the-path-to-ub… Demo chatbot: chatjimmy.ai API: taalas.com/api-request-fo…
Français

@Sword09D @KimNoel399 non mais discuter avec ce modèle on dirait un modele de 0,6b c'est claqué au sol.
Français

@ray5ar @KimNoel399 Je ne pense pas qu'il soit meilleur que ça honnêtement Q4 en 8b c'est vraiment le bas du panier
Français

@Sword09D @KimNoel399 llama 3 8b en q4 est moins con que ça. c'est peut-être une quantisation buggée?
Français

@ray5ar @KimNoel399 Apparemment c'est une version quantisée en Q4, ça explique que oui effectivement il est complétement con
Français

@KimNoel399 J'ai testé leur chatbot, le modèle est complémentent débile. C'est clairement pas llama3 8b je sais pas qu'est-ce qu'ils ont foutu mais c'est pas du tout bon.
Mais c'est clairement le futur pour réduire les couts de l'inférence basique.
Français









