Tweet ghim
Chetaslua
9K posts
Chetaslua
@chetaslua
AI News | AI Prompting and Comparison| Breaking AI news before it’s famous
Tham gia Aralık 2024
124 Đang theo dõi18.8K Người theo dõi
Introducing Benchmark Concept
Monopoly bench - i was thinking how to know which ai model is good in maths , planning and behaviour - for this only game is good
Thanks @Kimi_Moonshot @MiniMax_AI for the api ,
If anyone can help me to get @OpenAI , @claudeai , @GoogleAI api please text me i need them to run it on bench
> this monopoly has 4 characters - 4 different ai models
> bank is also ai model - you can chat with everyone in one gc - briber , partnership and win
> everything here is vibecoded in 2 hours so many things still missing
> i want to add more rules to it to make it more complex form of monopoly
one more thing this is my first attempt , so anyone who made benchmark for ai already can help me to guide or love to build it together i am open for all ideas
English
MiniMax M2.7 multi agents system
made a water physics with proper temperature control website
i have shared the codepen link and also @MiniMax_AI now please cook big model 1T or 2T , i have high hope from you guys
Chetaslua@chetaslua
MiniMax M2.7 is real deal man I have made a complete website with proper fan physics and fibre physics this looks so good @MiniMax_AI i am waiting for the next update man , i want a big model next cook it for us Prompt is in comment for normal paper receipt physics sim
English
@chetaslua @MiniMax_AI minimax doing water physics while my local multi-agent system is still hallucinating its own test cases
1t model gonna be crazy but deep architecture is the real final boss
English
@chetaslua @MiniMax_AI Man you are crushing it with these MiniMax builds! Keep em coming!
English
@chetaslua @MiniMax_AI vibe coder be like: look at my 3 FPS animation if I use a simple scroll bar. how awesome! guys!!!!
English
English
Google won't refund a single dollar on my $250/month AI Ultra subscription.
I paid for what I thought was a reliable, premium AI product.
Gemini 3.1 Pro ranked #22 on BridgeBench.
Gemini CLI crashes mid-session.
Antigravity can't hold context.
$250/month and they won't even prorate a refund.
Just "subscription ends April 12."
No refunds on a product that doesn't deliver is insane.
English
@HarshithLucky3 good job harshith , i am super proud of you and finally you getting love , you now know 👀
English
This new update met my expectations
Firebase backend integration works really well
> I tested it with a multiplayer 3D game. I shared the link with four people, and they joined using a game code. It worked perfectly
Yeah, this is a major update
I will test it more and share results in 1-2 days
Google AI@GoogleAI
We’re launching a brand new, full-stack vibe coding experience in @GoogleAIStudio, made possible by integrations with the @Antigravity coding agent and @Firebase backends. This unlocks: — Full-stack multiplayer experiences: Create complex, multiplayer apps with fully-featured UIs and backends directly within AI Studio — Connection to real-world services: Build applications that connect to live data sources, databases, or payment processors and the Antigravity agent will securely store your API credentials for you — A smarter agent that works even when you don't: By maintaining a deeper understanding of your project structure and chat history, the agent can execute multi-step code edits from simpler prompts. It also remembers where you left off and completes your tasks while you’re away, so you can seamlessly resume your builds from anywhere — Configuration of database connections and authentication flows: Add Firebase integration to provision Cloud Firestore for databases and Firebase authentication for secure sign-in This demo displays what can be built in the new vibe coding experience in AI Studio. Geoseeker is a full-stack application that manages real-time multiplayer states, compass-based logic, and an external API integration with @GoogleMaps 🕹️
English
@blackboxai @MiniMax_AI Damn bro saw your subreddit, it's all over the place
English
@chetaslua @MiniMax_AI MiniMax M2.7 is cooking hard. We'd say it’s even more powerful when you run it alongside other agents and compare outputs side by side. You should try our latest multi agent CLI if you haven't yet.
x.com/blackboxai/sta…
BLACKBOX AI@blackboxai
New release: Multi-agent part 1
English
MiniMax M2.7 is real deal man
I have made a complete website with proper fan physics and fibre physics this looks so good
@MiniMax_AI i am waiting for the next update man , i want a big model next cook it for us
Prompt is in comment for normal paper receipt physics sim
Chetaslua@chetaslua
My first test of MiniMax M2.7 this made 3d fibre physics with all the specs of m2.7 is written
English
@chetaslua the most underreported detail is that Kimi Claw launched and Moonshot's monthly sales exceeded their entire previous year's revenue. that's not a valuation story. that's a product market fit story. the $18B is just investors catching up to what the revenue already proved
English
🐐 Hot Kimi Take
People forget Nvidia lost $600B in a single day the last time a Chinese lab open-sourced something major.
Now Kimi is raising up to $1B at ~$18B. That's a 4x jump in 3 months making Moonshot the fastest-growing AI company in China.
They're open-sourcing frontier models and making inferences absurdly cheap.
All at 1% of OpenAI's valuation.
I don't know what else you need to hear.
Kimi.ai@Kimi_Moonshot
Introducing 𝑨𝒕𝒕𝒆𝒏𝒕𝒊𝒐𝒏 𝑹𝒆𝒔𝒊𝒅𝒖𝒂𝒍𝒔: Rethinking depth-wise aggregation. Residual connections have long relied on fixed, uniform accumulation. Inspired by the duality of time and depth, we introduce Attention Residuals, replacing standard depth-wise recurrence with learned, input-dependent attention over preceding layers. 🔹 Enables networks to selectively retrieve past representations, naturally mitigating dilution and hidden-state growth. 🔹 Introduces Block AttnRes, partitioning layers into compressed blocks to make cross-layer attention practical at scale. 🔹 Serves as an efficient drop-in replacement, demonstrating a 1.25x compute advantage with negligible (<2%) inference latency overhead. 🔹 Validated on the Kimi Linear architecture (48B total, 3B activated parameters), delivering consistent downstream performance gains. 🔗Full report: github.com/MoonshotAI/Att…
English
@chetaslua For me kimi is the new deepseek and I am ready to die on this hill
English
@chetaslua Kimi is my favourite prompt engineer, I still remember the day @ai_for_success post and after few days later @IamEmily2050 told me to use it ❤️
English
@Hibara_AI_lover Small model smell for sure , 10 b activated only , so you are right I pointed that in my post that's why I told them need big models now
English
i think we REALLY need to rethink about how much we trust @chetaslua's model test. Sure his/her (idk) leaks is good and i am not saying he/she is not a nice person. but minimax's models are NOT GOOD, TOO benchmaxxed and trained on these one shot test
Chetaslua@chetaslua
MiniMax M2.7 is real deal man I have made a complete website with proper fan physics and fibre physics this looks so good @MiniMax_AI i am waiting for the next update man , i want a big model next cook it for us Prompt is in comment for normal paper receipt physics sim
English
@chetaslua @MiniMax_AI M3 is already cooking, bro. 👀
English
@chetaslua its actually a lot funnier than that x.com/pigeon__s/stat…
ρ:ɡeσn@pigeon__s
wait this is hilarious i just realized grok-4.20's base model scores the same as MiMo-V2-Flash imagine this Grok model being hyped up and waited for for like 6 months straight is not even at its core smarter than a 200B Xiaomi model lololol its also less price and token efficient
English
Funniest thing happen while i was offline 😂
i just realized grok-4.20's base model scores the same as Gpt 5.4 mini
Imagine this Grok model being hyped up like AGI and waited for for like 6 months straight is not even at its core smarter
Proof in the quote last year finding
Chetaslua@chetaslua
We found this line in the Grok config, : “Grok 4.20 AGI (beta)” Artificial Grok Intelligence Probably just an easter egg… or maybe not 👀 @elonmusk can you confirm 🤯
English
@SNARKAMOTO openai has 600 million free user they cant give this big model lol
English
@chetaslua I don't have this info... I can't believe mini is so tiny. 300B+ at least.
English