Laks Dondeti 🇺🇸 retweeted
Musk insulting people is not new. But, if only insults cooled a data center in space...
Laks Dondeti 🇺🇸 retweeted
GrokBot was pitched as a colleague you could use for everything. People bought annual Ultra subscriptions - then the weekly limits were cut because serving it cost too much relative to revenue.
For me, I still prefer Claude for most tasks.
Laks Dondeti 🇺🇸 retweeted
FinalLayer is almost at $2 million in annualized revenue, all inbound so far. To test another growth channel, I picked four Instagram videos to boost.
Meta rejected my recap on the AI safety debate. I guess we cannot discuss facts in ads too now.
Is @OpenAI going be renamed OpenSI OR become THE ENEMY? 😭
This tweet is most likely false, but still funny? 😬
2
323
Laks Dondeti 🇺🇸 retweeted
Are you running a SICo?
Are you prominent enough to be part of the SIF?
For those not paying attention:
SICo: super intelligence company
SIF: super intelligence force
😭
Are you running a SICo?
Are you prominent enough to be part of the SIF?
For those not paying attention:
SICo: super intelligence company
SIF: super intelligence force
😭
Replying to @Subscriptions
Laks Dondeti 🇺🇸 retweeted
Replying to @Subscriptions
Laks Dondeti 🇺🇸 retweeted
Replying to @trq212
2 questions:
1) why is the world revolving around an interface from the 80s? More importantly, is that what Claude wants to showcase?
2) why do I have to update Claude code in the Mac app and at the cli level?
Laks Dondeti 🇺🇸 retweeted
Replying to @ajambrosino
A critical piece of the puzzle.
An app can’t become complex for its simple use cases no matter how many new features we add!
TikTok is an excellent example of this.
For some people, it’s just a feed of videos.
That’s the same experience for a new user a decade ago or a new user who first installs TikTok today.
Laks Dondeti 🇺🇸 retweeted
Replying to @sama
This is interesting.
LLMs are supposed to be smart and supposed to be capable of solving hard problems. There’s also some evidence of that.
Stuff we don’t like doing, the repetitive stuff and the mind numbing stuff, can be done with simpler models, including open weights models on local machines.
Curious why dot is your favorite product when there are ChatGPT and codex?
Just because we can build anything because we have LLMs doesn’t mean we build random things and shove them into a daily-use product!
@ClaudeDevs @OpenAIDevs you are on track to make Claude Code and Codex worse!
2
227
Tried “mod” on Claude Code.
It’s a confusing mess.
The tweets promised a lot! The documents are a mess and the whole thing is unnecessarily complex.
I’m surprised that Frontier Labs are consistently poor at FTUX.
I know what to do next, and I’ll report back.
But it should have been easier and more fun!
Tried projects on Claude.
Horrible FTUX.
Insists on migrating my old projects. If I don’t want to, I’m supposed to answer an MCQ.
And … done!
Wtf!
Anyway, I archived all my old projects.
Made a new project and pointed it at one of my “old” projects(naming from Frontier Labs leaves a lot to be desired).
Not bad! 😅
Tokenmaxxing … unlocked!
Tried “dot” on Codex!
Underwhelming.
I’m sure there are some use cases where it might be good.
I’m sure some will say that I don’t know how to use them! Please…
But the dot promised me that it’ll wow me!
Initial impressions: no wow factor.
Laks Dondeti 🇺🇸 retweeted
Replying to @thsottiaux
This is getting annoying. Banked resets are good, these random resets are very disruptive. It's like playing blackjack in a casino, house always wins.





