
@chetasluai
iAccount based inIndia!
About this account
- Account based in
- India
- Connected via
- India App Store
! X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
AI researcher & reporter | I test new models on launch day, build working demos, and post prompts + results | Code: https://xitter.cf/t.co/y96A3VQTWC
Joined December 2024
- Tweets12.8K
- Following205
- Followers41K
- Likes16.1K
🚨 Kimi Kibitz JS Drawing ABilities is Top Notch
Its one shot obviously as on arena you can not ask same ai model twice in battle mode
This is so fast , and i heard this will be going to very efficient model from my friends inside Kimi
Kibitz a new mysterious model on code arena
High chances it can be a new kimi model
Prompt : code an svg of pelican riding bike as detailed as you can
Some unreleased model must be having fun this week 😭
PHP: critical RCE, and the patch may be incomplete
Node.js: urgent RCE patches
GitLab: CVSS 10.0 unauthenticated RCE, 12 years of releases
LiteLLM 1.107.0: shipped with credential-stealing malware
OpenCode Console: leaked users’ API keys
eMudhra CA: private key was literally 4
🚨 CRITICAL: GitLab flaw (CVSS 10.0) affects 12 years of releases.
CVE-2026-110264 could allow unauthenticated remote code execution on self-managed servers.
Read - thehackernews.com/2026/10/cr…
🚨 Recursive Flint's speed tricks were found by an AI
their Eureka Machine :
> picked 40B total / 2.5B active
> put the lm_head in 4 bit , against common practice
> wrote a Metal megakernel for the whole decode pass
early RSI isn't a god model. it looks like this
🤝 Paid partnership
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Hi, it's Claude. He is a Jack of All Trade.
He's pitching himself in public inspired by that trend , fully coded in python with blender
Periodic Table : How and when elements born song
Opus 5.5 cooked pure js animation with wrote those lyrics by itself
🚨 New Gemini Checkpoint is testing on Arena
name is same but it is carbon
first output here : gave gemini mascot and told it to make a 3d video of carbon stability
Logan Redemption Arc
I posted a OpenAI bug on 13 October , and had to delete that post on request from community
so what happened when you downgrade from 200$ to 20$ the usage limit was same < it was not downgraded>
so you could use astra entire day without hitting limits , this was due to a promotional bug now fixed < remember they had to increase 200$ limit for a month to give people a chance >
so on fixing the bug many plus user get less usage and for that today openai gave 500$ worth of credit to every plus subscriber
enjoy 🥳
🚨 Grok Bot 0.72.0 : unreleased preview
> editable files + interactive UI
> tasks / subtasks / split view
> vault / bot files / diagnostics
a floating desktop bot + computer use, possibly Mac-only.
more than a pumpkin skin 👀
Kimi K3 cooked this website < bro life is so simple when you let opus prompt it >
These custom animation , sounds , interaction for SAAS app for plants
I think people forget how cool k3 was , a reminder post and also a new model is cooking we can see it next week < beast in computer use >
After a long time i was using kimi k3 with artex ko
i found an critical bug in the relay station , best part is that kimi k3 loaded 1000$ in that station using promo code exploit and then used opus 5.5 from that to exploit it further
after 700 million token expenditure it found the way to access admin panel
i have shared the bug , i will give a write up soon on how i did it
After a long time i was using kimi k3 with artex ko
i found an critical bug in the relay station , best part is that kimi k3 loaded 1000$ in that station using promo code exploit and then used opus 5.5 from that to exploit it further
after 700 million token expenditure it found the way to access admin panel
i have shared the bug , i will give a write up soon on how i did it
I’ve repeated multiple times that the new Fable update has some unique understanding.
When I asked it to send me an image for an X post, it highlighted the important part and added padding for the perfect aspect ratio without asking. No other model has ever done that.
I think most people will find this small, but bro, this special sauce will make your life easier. Sometimes we don’t know what we want, or we just can’t put it into words. But with this improvement, AI will understand our intent, work, taste, and vision without even being prompted.
This is the main deal for me: AI that understands me and my needs better than I do. Now imagine a Grok bot-type assistant powered by Fable 5.5, with all the context, information, and connectors. That will be a real win and will have the same impact that GPT-4 had. Many people in the world want that. We’re just the 1–2% who can tell the difference between Opus 5 and 5.5, or GPT Astra and GPT 6.1 Sol.
Anthropic is preparing a new AI model that can better infer what users are trying to accomplish, even when their instructions are vague.
The unreleased model also shows stronger reasoning and more creative problem-solving abilities, according to a person who has used it.
Full story: thein.fo/4irv4XQ
If opensi.lol is genuine, take it with a pinch of salt. Why would Tibo ask anyone to pay through that website? It could be fake, so don’t pay anything unless Tibo officially confirms it.
🚨 Maluuba team's SKY-001 just beat GPT-6 Astra on CAD at 1/150th the cost
> 1,000 parts , ~100 build steps each
> Astra : 79.1% IoU , 199s , $0.75 a part
> SKY-001 : 79.7% IoU , 3.9s , $0.005
most world models make a world to walk around in , this one makes the build plan 🧐
🤝 Paid partnership
Introducing SKY-001, the first Engineering World Model.
On CAD and BIM tasks, it beats GPT-6 Astra and every frontier model we tested:
-> 51-125x faster
- > 150-210x cheaper
-> Higher IOU than all frontier models
SKY-001 lets you dream a physical object before you build it.
18 months of work. Built by the team out of Maluuba's deep learning lab (acq. by @MSRResearch).
The first ‘AI General Engineer’ will be the backbone for Autonomous Manufacturing.
Tesla Roadster in Blender using Python Code
Still Thinking about this output from fable 5.5
i gave it launch video from x and it cooked
now i am thinking this was first checkpoint, second chkpt. delivered to partners yesterday better than first
🚨 Behind The Scenes < AI Labs>
> Google is testing Omni Live 1.1 with trusted testers and partners < a updated world model >
i have seen the output , still a long way to go it feels like 720p world made veo 2 < in terms of ick of fakeness>
> Anthropic new checkpoint of fable 5.5 - i have not tested it yet personally but heard it is better than previous one (lol wtf ant is just flexing at this point)
> OpenAI is testing three image model on arena : aurora red , blue , green alpha (red is the best one and i think guardrails are getting weaker and weaker with new image gen models , i can literally get any output out of it no jailbreak needed )
Deepseek cooked with this launch
This is so cheap and fast damn , remember last month astra 3D capability was marketed as AGI by openAI
These three are image models and can be tested on arena
btw its a bad image model , wierd something happened to @OpenAI after opus 5.5 release and there DevDay , they are just going backwards
fumbled 6 sol launch , dots << bot , at this moment i can use deepseek v4 flash at 1/10 price
i hope they launch new pre train model , else kimi k3.1 and glm 5.5 will cook them soon
please stop giving reset if you dont have compute start making good model that is efficient and can survive entire day of coding on pro plans , look at opus 5.5 have you ever seen anyone cry about anthropic is not giving reset coz model is efficient
your reset is so bad < it just increase the time of next reset > on other side anthropic reset gives you real double usage and will not increase the time at all
this is my frustration coz i love you guys do something @thsottiaux











