Rohit Tidke

The answer to the meaning of life is the Universe and asking the right questions is our purpose!

[Rohit Tidke @_rtidke]

@elonmusk @SawyerMerritt So I had more than a month ago predicated this, proud of myself.

[Rohit Tidke @_rtidke]

@stalman That is going to cost a lot, 1.5 billion iphones, no wonder they support only 15 pro onwards, even for that its going eat a lot of money, so now the iphone 17 pro is a good deal bang on the buck, this is almost minimum 100€-300€ of compute per iphone if super actively used

[Rohit Tidke @_rtidke]

@mweinbach @stalman No idea what those words mean, copy Gemin base model RLHF on Apple stuff?

[Rohit Tidke @_rtidke]

@mweinbach @stalman what does that mean? how can you have 2 llms running?

[Rohit Tidke @_rtidke]

@munster_gene what you mean its not yet launched???

[Rohit Tidke @_rtidke]

@Bitcoin_Teddy Lol Markets do not just fall, the falls are coordinated, its common sense that people invested are invested cause they basically need the liquidity at some point, the issue is they do not know that point, and this is the biggest leverage of coordinated selling.

[MiniMax (official) @MiniMax_AI]

$0.07 for M3, $3.39 for Opus. Both caught 13 of 17 bugs.

Really interesting breakdown from @kilocode

Definitely worth the read

[Rohit Tidke @_rtidke]

@jun_song Ok then why do they need to run their AI workloads on google cloud using Nvidia chips in 2026

[Rohit Tidke @_rtidke]

@kimmonismus This happened with ChatGPT too

[Rohit Tidke @_rtidke]

@jawwwn_ and they went to Anthropic

[Sumit Kumar @TweetsOfSumit]

Hetzner should become Europes sovereign datacenter & cloud powerhouse.

[MiniMax (official) @MiniMax_AI]

15.6× faster decoding at 1M tokens 🔥

Thanks @FireworksAI_HQ for powering the inference behind M3.

Try it now 👇

[Rohit Tidke @_rtidke]

@StockSavvyShay He lives in some other world

In his world AI is creating jobs
In my world young students are suffering unable to find a position for years

In his world ROI is high, all I know its Megacaps making money, rest most companies are on the losing spree.

[Lord Bebo @MyLordBebo]

🇩🇪 A guy has to pay one monthly salary for calling Chancellor Friedrich Merz “Lying Fritz”.

The court ruled that saying “Lying Fritz” on Facebook is an insult that cannot be directed at politicians.

The exact fine amount is unknown. The court sentenced the “dangerous criminal” to 30 daily income rates, equivalent to one monthly salary. So he has to pay whatever he earns in a month.

So if the person earns $5000 a month, that’s the fine.

[Bridgebench @bridgebench]

GPT 5.5 limits have been significantly reduced.

[Rohit Tidke @_rtidke]

@bridgemindai super! looking forward

[BridgeMind @bridgemindai]

MINIMAX M3 JUST DROPPED.

1M token context.
$0.30 in, $1.20 out.

It even takes video as input.

This is not a small update.

It is a brand new architecture built for speed.

The claim: 15.6x faster at a full million tokens than the last model.

If that holds up, it makes giant context AI agents cheap enough to actually use.

But speed is not quality.

That is what the BridgeBench is for.

BridgeBench results and my full review drop tomorrow.

https://t.co/ttnnwBYerW

[Tibo @thsottiaux]

The Codex usage limits have been reset for all paid ChatGPT subscriptions. You should be back to 100% weekly and 100% hourly limits.

Let the tokens do incredible things today and have fun.

[Rohit Tidke @_rtidke]

@thsottiaux @QiaochuYuan I think the major diff is speed, OPUS gives the illusion or is faster on claude code, CODEX feels slower, even though it could be effectively even faster, this is some thing nuanced, cannot explain why

[Rohit Tidke @_rtidke]

@DavidKPiano saving context right

[Rohit Tidke @_rtidke]

@kimmonismus ohhh no, Opus models I just think they go worse in some aspects on each mini iteration ,sonnet 4.8 would have been much much better !