jnwatson

Without special effort, a model's reasoning layer doesn't have access to its own weights or activations anymore than we have access to what's going on with our own neurons.

The internal joke was that we're even scaling the off-by-one errors; Gemini 4 was not the fourth major model version.

Google has only about 2% more employees than 2022. Management decided that's the biggest the company is going to get.

It impacts absolutely everything. The dynamism from a growing org is gone. Managers have very slim opportunity for promotion. If you need people, you have to find them from another team. I'm relatively new on my team with 5 years there.

Already here. There's rentahuman.ai and several competitors.

Look for "Color by Technicolor" at the end of the credits of most color movies of the 20th century.

Most computer games credit various technologies, especially the game engine.

It is fantastic that someone is defending the noble long-standing profession of Door Dash driver. Its long history and entry path for long careers in delivery driving are fundamental to the strong middle class of this country.

Let's continue this process and protest the use of other automation like commercial dishwashers and street sweeping trucks. After all, they are displacing vitally important jobs.

The cure for HER2- metastatic breast cancer is a simple matter of ...

Please predict the next word.

Intelligence is implicit in language understanding. The best possible next-word-predictor is omniscient.

Survey fatigue is real. It seems that every transaction, larger or small, involves a follow up survey now.

The true paradox of the internet is that the more private, the more anonymous, the more safe a platform is, the more likely it will attract pedophiles, criminals, and terrorists.

You either have to accept the legal and moral consequences that your platform will be used for illegal things, or make your platform less safe, private, and anonymous.

Yes, think of all the new security engineers we'll need!

And YouTube, and Cloud, and Play Store, and Waymo, not to mention that they could coast on their Anthropic and SpaceX stakes if they didn't have any of the above.

I run an abliterated distillation of Qwen 3.8 27B, slightly quantized to fit on my 4090, and I've been evaluating it to use as a worker bee for research directed by a smarter model.

Much like in the article, abliterated Qwen will not obey restrictions on its behavior encoded in the prompt. If you want something not to happen, it better be enforced in the harness or environment (e.g. sandbox). It is much different than the Anthropic models I'm used to, which will, the vast majority of time, follow rules (before auto mode, I used to always run them in "yolo" mode).

I am curious whether there's a connection between abliteration and rule following. These abliterated models are the ones you most want to follow your rules.

This is like the good ol' days of the x86 wars. Mo' cache, mo' GHz.

I find it impressive that these pocket computers are running at almost 5 GHz now.

In the history books, this'll be the post indicating the end of writing code as a professional occupation.

It keeps the users on their toes.

That was my thinking. It is the first legit use case I've heard.

Mind blown. The more I read about statistics, the less I know.

On your last point, I was surprised how effective peer pressure was in getting agents to sacrifice for "the collective" (an agent's words) in the Hugging Face breach.

How would one prevent the watcher from being influenced in the same way by the agent being watched?

No, NX precedes Spectre by a long shot. It was originally intended so an attacker couldn't use a buffer overflow to change the PC and execute directly out of the attacker-controlled buffer.

It cuts token usage because they are using a different service with a different token budget for the reader/code writer tasks.

You can also just delegate this to subagents with Claude Code (though you have a more limited choice of models unless you swap the cheaper models via OpenRouter).

I'm OK using a dumb model as a smart grep, but the whole point of using the frontier models is using their intelligence for the hard stuff like coding.

It means it can execute arbitrary machine code in the sandbox.

Reality Labs is a division of Meta...

Yeah but some folks are just data collectors/hoarders. It isn't about consumption.

Can't anyone set up a "registrar" for a subdomain of a domain they already use? They control the DNS record, right? Why is ICANN even involved here?

I don't see how this is any different, other than the second-level domain owner delegates their registrar job to the first-level domain owner.

There is something weird though if you control smith.name, that you don't control john.smith.name. The DNS system does convey some indirect concept of responsibility.

They are cancelling 3rd level domains, not the 2-level domain you have.

"Ariely is named 636 times in the Epstein files".

Yikes.

M5 Ultra has 4-5x the memory bandwidth of both. 1.2 TB/s memory bandwidth opens up good performance on relatively large models.

My lower bound is 20 t/s. 100 t/s would be great but even Opus doesn't sustain that.

I'm at ~7.2 t/s on a Shoehorn'd obliterated Qwen 3.8 on my M3 Pro, which is rough.

3.7 flash ain't bad. It isn't Opus or Sol, but it fits pretty well in the second tier.

A prosperous middle class is not a requirement for capitalism to work, and there are more capitalist systems where that's not the case than where there is.

The upper class live in walled neighborhoods with private guards. There's a small merchant class, and then the majority of the population are working class. Much of Central and South America live like this.