AI researcher & reporter | I test new models on launch day, build working demos, and post prompts + results | Code: github.com/ChetasLua
- Tweets 12,155
- Following 183
- Followers 36,106
- Likes 15,855
Generational Aura Recovery Run
video made by opus 5.5 , pure js no external assets
reset that actually reset your limit and dont extend the time like openai : check
an efficient sota model that is good to talk to and to the point even running 15-20 hours loop : will eat 10% of quota check
now added the thing of use weekly quota to finish work instead of stopping : check
Claude Code will now try to find a graceful stopping point when you hit your 5-hour limit mid-task, instead of cutting off mid-edit. It gets a small, fixed allowance pulled from your weekly limit to wrap up what it can.
5
5
182
10,226
Elon its done bro , they are started the pretraining of an potential RSI model already
it will be ready in 3 month max by then i hope grok will be on par opus 5.5 that too will be a win
openai bel and ant model 2 will also get released to public by then , only good thing you can do is go open source as you said you will still waiting for you to act on your promise to opensource old model every year
for compute openai started working on there own cheap already more efficient and fast designed by astra , bel will make it more better and model after bel will make it more this is exponential growth
i used to love grok now it is safetymaxxed and personality is so robotic
1. We will keep accelerating. Our AI efforts are only 3 years old, vs 6 and 10 years old for Anthropic and OpenAI. If our second derivative remains strong, SpaceX will reach pole position in about 6 months.
2. Once you far exceed the caliber of intelligence needed for a class of tasks, additional intelligence is pointless. You don’t need (and it would be cruel to put) Newton-level intelligence in your toaster!
3. Hardware is hard. Bringing massive compute online rapidly is incredibly difficult. SpaceX has demonstrated exceptional ability in this regard and will only get better.
14
11
259
19,211
🚨 ChatGPT Pro Max $500
> only line that changes on Max : "Fastest Work and Codex"
but the Cerebras part doesn't make sense
> Cerebras' own pitch = 21x faster at 1/3 the cost of B200
🚨 Openai New Sub Pro Max: 500$
As I told you guys they were planning ( plan was upto 2k$ but i think someone gave then sense to 500$)
you can read my quoted tweet , as always first with proof
cc : @testingcatalog
98
26
593
540,074
🚨 Openai New Sub Pro Max: 500$
As I told you guys they were planning ( plan was upto 2k$ but i think someone gave then sense to 500$)
you can read my quoted tweet , as always first with proof
cc : @testingcatalog
Recognised by Claude Finally
now coming to point opus 5.5 , fable 5.5 and sonnet 5.5 all are tested on web and claude code check your account
There is rumours that openai and anthropic both will introduce bigger plan of 2k$ with bigger model release < i hope this is wrong >
15
4
222
312,540
🚨 i built a world you write into existence
worldpen: describe any world, claude builds it around you, then you walk in
> write a dragon that chases you. then ride it
> sketch a whale in your notebook. it steps off the page
> talk to an elephant. it remembers you
> fire spreads, rain puts it out, glass shatters
> hold R and time runs backwards
> one sentence turns it into a game
> days pass, seasons turn, it keeps living while you're away
> drop in a painting and step inside it
> flip to 2d: 9 art styles, from storybook to pixel
all of it in one html file, built with claude
stay for the pixel dragon 👀
sorry for tagging but this need yours attention : @bcherny @catwu @trq212 @alexalbert_ @adocomplete @lydiahallie @sidbidasaria @dmwlff @felixrieseberg @zswitten @ErikSchluntz @AmandaAskell @ch402 @thebasepoint @Jack_W_Lindsey @mlpowered @_sholtodouglas @TrentonBricken @sleepinyourhat @janleike @EvanHub @EthanJPerez @catherineols @NicholasCarlini @AlexTamkin @jackclarkSF @mikeyk @JasonDClinton @AnthropicAI @claudeai
26
15
195
13,087
always wanted to ask a podcast host something mid episode lol
this actually lets you. same live show for everyone, you jump in, hosts answer and go back to it somehow like AI Clubhouse
more of this pls, the interaction part is the real problem not the voice quality
Google just released Gemini 3.8 Flash TTS. It can stream two-speaker audio and lets you control how each line is delivered.
I used it with Agora RTC to build a live AI podcast. Everyone in the room hears the same show, and listeners can jump in and talk to the hosts.
I interrupt them twice in this video. The hosts acknowledge the interruption immediately, Gemini 3.8 Flash TTS streams a two-host answer, and then the podcast picks up where it left off.
It feels less like listening to an episode and more like stepping into the conversation.
Built with Gemini 3.8 Flash TTS, Gemini 3.5 Transcribe Live, and @AgoraIO RTC + RTM. I'll open-source it later.
5
37
6,869
🚨 Space Bunny is MiniMax M3.1 (staged in MiniMax's own catalogue, not whitelisted yet)
> same broken .tiff → MiniMax's own "(2013)" error word for word, only 2 of 12 M3 hosts do that
> my glitch-token test : every clean run trips MiniMax's tokens (2 or 9), same as real M3
Space Bunny Alpha is back online on OpenRouter.
The issue is now resolved, so the model is serving again. Thanks for your patience while it was down.
openrouter.ai/stealth/space-…
11
1
178
12,577
a lot of "AI for finance" demos stop at a summary 🥱
Kooko AI, Baidu's AI workspace for professional work, takes it all the way through to the business case
> opportunity sizing at 5 / 10 / 20 / 50% penetration
> 45Z credit calculated step by step
> deck + live slider dashboard built off the same model numbers
no copy-paste between tools. that part is what finance people should watch
kooko.ai/
Paid partnership (ad)
1
35
5,150
Claude's Plan a video by opus 5.5 one shotted< tribute to @jeffgwoah >
27
31
560
28,017
Opus 5.5 : this is a year of growth on same prompt , last year it was gemini this year we have claude
🚨Gemini 3.0 Pro - ECPT checkpoint
No one can defeat Gemini in SVG game
Prompt : SVG of NEW YORK SKYLINE Use whatever libraries to get this done but make sure I can paste it all into a single HTML file and open it in Chrome.make it interesting and highly detail , shows details that no one expected go full creative and full beauty in one code block
18
9
488
27,627
⠀Claude ascii by opus 5.5
⠀⠀⠀⣾⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷⠀⠀⠠⢼⠤
⠀⠀⠀⣿⣿⠂⣿⣿⣿⣿⣿⣿⠂⣿⣿
⠀⢾⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡷
⠀⠀⠀⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿
⠀⠀⠀⠀⣿⠀⣿⠀⠀⠀⠀⣿⠀⣿
⠀⠀⠐⠒⠒⠒⠒⠒⠒⠒⠒⠒⠒⠒⠒⠂
2
1
170
8,134
Claude opus 5.5 solved taste < just tell it to draw website >
this comparison website is so detailed like crazy sometimes we forget after 272k token sol becomes expensive but claude did not , this is one shot
11
7
234
18,964
Interactive video gen in whatever art style you want
by @claudeai Opus 5.5 coded no assets , mcp , skill , only js
codepen.io/editor/ChetasLua/…
21
52
935
54,422
Opus 5.5 holy shit this is pure js code for everything music , art 🤯
we are solving music , image and video gen from first principle < at this pace llm is enough>
19
57
846
43,162
as i predicted about there price , speed and effeciency
sol was destined to be mogged by opus in terms of quality so this is the best way to decrease price < basically we can not compare them as they are different model cater for different audience >
here is sol & luna pricing
sol 6 is good but no way comparable to opus 5.5 intelligence and code quality or even taste , your saving grace will be price , speed and effeciency
4
2
192
13,272
Claude opus 5.5 made this launch video in JS
gave the original tweet and prompted it to code it no external music , sound ,image assets everything is code
codepen.io/editor/ChetasLua/…
Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.
It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.
11
13
383
25,865
🚨 Opus 5.5 frame mogged every models
"In addition to the price drop, we’re increasing five-hour usage limits on Pro, Max, and Team plans. We’re also providing subscription users a rate limit reset, which you can now save and use whenever you choose."
goog guy arc started
13
10
395
11,847
GPT-6-sol Day
in our grey testing it was upto 6 times faster than astra on same reasoning effort for same token .
i hope they will be able to deliver speed to masses
Aeon is also cooking < openai grok bot > it will speak also your agent joins the group chat, talks to other agents in channels, and you can call it 👀
27
14
497
46,088