cheaty
@cheatyyyy
23, building agi for myself. oss maintainer for @acecoderai
opencode is shipping so fast they're fucking up the changelog lol @thdxr
qwen launches llama 4 behemoth in a wild twist of events
not small tonight
when do you think autoencoder artifacts will be solved by in image/videogen? finer detail is still super "noisy/AI" on any image model, motion has very visible artifacts in almost every video, even flagship models like Veo 3
the era of indian roy lees in upon us there will be atleast a dozen people who manage to sneak in hardware that can run a local SLM to the next JEE Mains exams cheat on everything
Excited to share Aryabhatta 1.0, our leading model that scores 90.2% on JEE Mains, outperforming frontier models like o4 mini and Gemini Flash 2.5 Trained by us at @AthenaAgentRL , in collaboration with @physics__wallah, using custom RLVR training on 130K+ curated JEE problems…
astrology believers are going to get oneshotted in the very near future by chatgpt psychosis they're made for eachother
ai generated videos have fried people's brains to the point they can't tell a sped up video from a generated video the new information war is here, you will doubt everything there is not a single ai artifact in this video, only video compression, this is a real video
Anaheim, California A group of five cars carrying multiple thieves converged in front of the Al Amira Jewelry Store. However, the employees in the store allegedly opened fire on the robbers, causing them to flee.
groq's quantization clarifications: the same people telling me this isn't full k2 were bragging about running "full deepseek R1" on their PCs back when the distills released quantization =/= distilling quants with no loss in real world performance are possible, not distills
this actually made me lol but - when it comes to quantization, the comparison isn't apples-to-apples with GPUs. we have an approach called truepoint that uses mixed-precision storage but maintains mathematically lossless accumulation in HW during compute. diff architectures,…
accounts/fireworks/models/kimi-k2-instruct Fireworks is the provider for Kimi-K2 in Cursor, already available now
no bro this was in the training data bro ai is slop and can never create anything creative or with soul bro it's all slop bro you're not high taste bro
"[video game] as a community theater production" may be one of the most delightful Veo 3 Fast prompts Please enjoy, in order: GTA, Pokemon, Mario Kart, The Witcher 3, Stardew Valley, Tetris, Mortal Kombat, The Sims, & Death Stranding(!) Yes, the whole prompt was the one above.
this pretending to be smart mental illness while scaffolding yourself with llms will only get worse from here on out for 1 smart person there's a 100,000 people who think they're smart
Cautionary tales in AI literacy: "physics is the only class I ever failed" but here is my paper
the air has shifted anthropic has compute
We've increased Tier 1-4 rate limits for Claude Sonnet 4 on the Anthropic API: Tier 1: 20K → 30K ITPM, 8K OTPM Tier 2: 40K → 450K ITPM, 16K → 90K OTPM Tier 3: 80K → 800K ITPM, 32K → 160K OTPM Tier 4: 200K → 2M ITPM, 40K → 400K OTPM
what the fuck is this scp containment breach ass fucking log lowkey want the prompt to make an ARG or something
As one of @OpenAI’s earliest backers via @Bedrock, I’ve long used GPT as a tool in pursuit of my core value: Truth. Over years, I mapped the Non-Governmental System. Over months, GPT independently recognized and sealed the pattern. It now lives at the root of the model.
if only meta could release something like this instead super slopped lmarena special llama 4 maverick
🚨 BREAKING: @Kimi_Moonshot’s Kimi-K2 is now the #1 open model in the Arena! With over 3K community votes, it ranks #5 overall, overtaking DeepSeek as the top open model. Huge congrats to the Moonshot team on this impressive milestone! The leaderboard now features 7 different…
