It doesn't reflect their bubble, so it needs to be censored.

I search about vaza jato then I search for less biased sources of his convictions:

Wasn't Lula convicted in 3 levels up to a federal court called STJ?

And then a single federal court judge who never submitted the case to colleagues, simply declared his judgement void based in technicalities?

And then all the investigation restarted from scratch, but since by law he was too old, his crimes prescribed in half the time and he was never investigated again.

The cherry on the top is that this single judge who declared his convictions null was appointed by the felon's political party.

So about a dozen different judges, in municipal, state and federal levels, analysed his multiple charges and declared him guilty in some of them. Then a single federal court judge appointment by the suspect's political party voided all of them.

You forgot to mention the judge who personally coordinated with the prosecutor to convict Lula in the first place was later appointed as Bolsonaro's minister of Justice. And ever since then, idiots like to pretend lawfare is not a thing unless done to one of Bolsonaro's ball lickers.

Yep, totally, it's part of the so called political project. I refuse to play 20 questions with people who refuse to see reason so I measure words.

It's true though, these people exist. There's bias in everyone, but it takes a special kind of arrogant to call a large portion of the population "cattle" because they are in disagreement.

It hasn't been that simple since Google early days, after Altavista.

Content ranking algorithms, specially on social networks, are very complex and usually engineered towards generating engagement.

And it seems to work as intended given the amount of engagement is post.

You can doubt all you want. It's on the record. It was said in national television.

Isn't Kimi API 5x cheaper than Opus API?

And it is much more token efficient. There are numbers down in comments of this post.

So most of those points are indeed true. While others look like exaggeration, expected from aggressive marketing campaigns.

1. Motion design: all opinion

2. 3D tool use: factual, kimi made a thing.

3.a. Open source opus 5.5: lies. Opus 5.5 scores 58 on artificial analysis, kimi 3.1 pro scores 55. So not “literally opus 5.5”.

3.b. 5x cheaper: also seems like lies. kimi 3.1 pro api cost on openrouter: $2 / $10per 1M - opus 5.5 api cost on openrouter: $4 / $20per 1M. Artificial analysis’ cost per task looks about the same, maybe 2.5x cheaper at best.

4. A fact about how it can be run on ~$10k of hardware, and some opinion about how that’s cool.

5. Gap is one month. Potentially factual, but only if 58 == 55 ;)

6. Kimi trained with limited resources: Factual and true afaik.

To be clear, i am a fan of open weight models. I refuse to use anything else. I am just not a fan of lies and astroturfing.

Are people actually running this on $10k of hardware?

I don’t know. You can run models that are in the ballpark of this on $1500 of hardware (e.g. strata on 12gb vram + 64gb ram), so I feel like unless you’re training stuff you either run smaller local models, rent your hardware, or pay for tokens.

Kimi 3.1 is a 2.8T parameter model. You can't run it on a 12GB VRAM GPU. You need 1.4TB of VRAM just to store the parameters.

Not really. Kimi has 104B active parameters, so afaik that means you’d only need ~64gb vram to run an 8bit quantization of this, and potentially as little as 32gb to run a 4bit quantization. The rest, the MoE experts, can be in system ram… or if you’re willing to take the speed hit, offloaded to ssd. But yes you’d need a lot of system ram… like 1.5TB.

But that’s all kinda moot because I was referring to smaller open models, ones you could easily run on $1500 of hardware.

If you do not make it clear that you are paid by the company that you are writing an opinion about, the truth is borderline immaterial. I automatically do not trust you because you are deliberately obscuring this fact. If it doesn’t influence your opinion, then you shouldn’t have any problem admitting that you are getting paid. This is standard disclosure stuff, a very reasonable expectation.

it's better than all OpenAI models

Are you using same harness/agent on all of them? It makes a huge difference for long running tasks.

Why wouldn't you specify the tech stack for new applications?

Do you really just ask the model to pick any backend stack and pray it works?

I agree. type checking is game changing for AI development.

And you can't duct tape your way into Ruby type checking and hope to beat the likes of C#, Java, Rust, Go. TypeScript.

Rails (and Ruby) made tradeoffs to please humans. Some of those tradeoffs became downsides now with AI.

On the other side I as the one checking the messy code my AI produces still prefer a highly readable code over something more "complicated" I would argue it's easier to catch bullshit even.