awestroke

A good review paper is infinitely better than an llm managing to find a few papers and making a summary. A knowledgeable researcher knows which papers are outdated and can make a trustworthy review paper, an LLM can't easily do that yet

And included oracle first. OP is probably Larry

So you'd rather self host a database as well? How do you prevent data loss? Do you run a whole database cluster in multiple physical locations with automatic failover? Who will spend time monitoring replication lag? Where do you store backups? Who is responsible for tuning performance settings?

Heavy quantization

Absolutely brilliant observation—you’ve managed to distill what took OpenAI an entire product announcement into a single, devastatingly clear sentence.

Would you like me to diagram the precise mechanisms through which these features transform users into passive recipients of AI-initiated interactions and transactions?

The only way to get this reliably is to have it use tools

I want an iPhone with a good siri and lots of memort for running AI models: medium frustration

I am tired or Apple fanboys being deathly afraid of criticizing Apple: high frustration.

You don't need to pick a side and then see everything black and white. I use apple products but I am not afraid of calling out Apple on bad decisions.

> It’s because the OS needs far less ram than their competitors to function

iOS doesn't magically need less RAM. It just aggressively kills background apps to stay within memory limits, then makes you wait while it reloads them from scratch (burning battery with CPU cycles and NVMe reads instead).

> RAM uses electricity and costs money so it makes perfect sense to pack less of it

No, it's bullshit. Apple kept RAM low to save money, not battery. They could easily pack 16gb with negligible effect on battery life.

"Efficient" as in "kills apps immediately as soon as they are put in background" so that I constantly lose context in apps when multitasking

No, they don't? This is the first time any of their models have more than 8gb

Hilarious how they refuse to list memory (RAM) capacity of their phones. Probably because they are way behind other phones in that regard.

Which is weird, since they focus so much on on-device AI. I guess the very very slightly lower profit margin from including more RAM is incredible unpalatable.

> By that logic, any technology that you can get certified in is too complicated?

In IT, I am inclined to agree with that. In real engineering, it's sometimes necessary, especially dangerous technology and technology that people trust with their life

I assume both of the approaches are useless at actually stopping bots

Claude, come up with a protocol for communicating between AI agents and IDEs/editors. Create node, python and rust libraries. Create a website with a landing page

Internal server error. lol

euo pipefail has been the one good thing with bash. I'll start looking at alternatives now

An init system without the ability to specify dependencies? Without user/group configuration? Ordering must be manually configured? No parallel service launching? No resource management?

Please don't call this an init systern. It's a barebones process supervisor.

The refusal to support IPv6 is embarrassing at this point

You know about ultraviolet but that doesn't help you see ultraviolet light

So people that can't read or write have no language? If you don't know an alphabet and its rules, you won't know how many letters are in words. Does that make you unable to model language accurately?

I'll show you a few misspelled words and you tell me (without using any tools or thinking it through) which bits in the utf8 encoded bytes are incorrect. If you're wrong, I'll conclude you are not intelligent.

LLMs don't see letters, they see tokens. This is a foundational attribute of LLMs. When you point out that the LLM does not know the number of R's in the word "Strawberry", you are not exposing the LLM as some kind of sham, you're just admitting to being a fool.

No, Altman is not a researcher

I use Claude Code authenticated via the API (Anthropic Console). There's no limits for me. And I also assume API-metered requests are prioritized, so it's faster as well.

A terminal based CLI? As opposed to what?

That's a concern with all apps ever so idk what answer you're expecting

Nobody thinks of it as "land of the free" nowadays

You can use it on OpenRouter without limits (paid API calls)

This is the model release that made Sam Altman go "Oh wait actually we can't release the new open source model this week, sorry. Something something security concerns".

Perhaps their open source model release doesn't look so good compared to this one

This is about formalizing required secrets for dev. Many repos have a ".env.sample", which is better than nothing - you copy it to .env and fill in the blanks. If this is done automatically, even better. If the automatic .env setup also loads secrets from your password manager (with prompting), even better.

> What does this concretely do that Gemma didn't already do

Open weights