jral

Political views: All black anarchist (as in far/post left, no adjectives) Hobbies: Asking, "Why?", and often getting depressed about the answer. 😷✊🏴

[jral @jral]

People that talk about their COVID protection journeys remind me of kids on the court pumping their Reeboks thinking it’d give them more game. For the love of… go get fit tested, and stop following trends on social media with mask brands and models. Everyone’s head is different.

[jral @jral]

Moonshot being a ChatGPT 5.6-Sol rival doesn't impress me any. Sol can't follow instructions, and will do it's own stupid fucking thing and burn through tokens like no other. Just like GLM-5.2, just like Fable. Let's see it do actual brownfield work without breaking everything.

[jral @jral]

Every time humanity sees something they caused, "This is so dystopian". Too funny.

[jral @jral]

The only kind of news I want to hear about politicians is obituaries at this point. They're all feckless, worthless turds that shouldn't hold the office that they do.

[Leighton 明 Woodhouse @lwoodhouse]

Shit is all fake, end of the day they’re all on the same team.

[jral @jral]

The saddest thing to me about vibe coders is that these LLM in vendor official development frames can't even get grep syntax right, and people are actually building entire applications with them. The world is fucking doomed.

[jral @jral]

@cremieuxrecueil Oh my, so basically it's behaving like Gemini 3.1 Pro does.

[jral @jral]

Saw someone in Gyro Bites tonight that was masked (though it wasn't an N95) and I thought, wow, cool, she must care about her health. Until she got her food that is, which she took off her mask to eat at a table. What was the point?? This is why people can say "masks don't work".

[Justin Amash @justinamash]

Here’s the roll call of members of Congress who betrayed the Constitution and the American people by voting yes on the KIDS Act.

The sickest part is that these people used “protecting kids” as an excuse to further insert government between parents and their children—all while establishing a surveillance infrastructure to monitor and control Americans of all ages.

[jral @jral]

Gemini 3.1 Pro won't even analyze it, due to it being a crack. It's a simple task, where does this binary write to, and what -- that's not a harmful thing. Software archaeology involves a lot of pre-cracked binaries, reversing the patch is critical for byte-level analysis.

[jral @jral]

I keep seeing all of these reasoning AI touted as "Mythos-level" at security and exploit analysis, yet if you hand GLM-5.2 an old DOS MZ binary that writes to two places (a crack), it'll get duped by a switch and strings, and be unable to focus on finding the simple INT 21h call.

[jral @jral]

There is only one function in this entire file that does any I/O. This is something a human can do in Ghidra, in about 2 minutes. I don't think we're "there yet" for any form of serious security analysis from LLM.

[jral @jral]

I just love how service have "prime time" billing hours. Electricity, water, AI... You get billed more, for having the service and making use of it, and if your hours just so align with a bigger billable, oh well, that's life. It's total capitalist bullshit. Scale bigger humans.

[jral @jral]

@DanBochman For me, it's plan out a feature addition, look at what systems it'd touch, explain the entire thing to Opus 4.8 on High reasoning, tell it to do it after it SOUNDS like Opus understands, and then spend an hour on berating it, reverting the changes, and doing it myself.

[jral @jral]

Is it smarter than Opus 4.8? A little. Does it provide better code than something like GLM-5.1? Sure does, it spends a lot of time thinking, it's smart. But it's also just another card trick, rather than a junior engineer that can work with you properly.

[jral @jral]

Fable 5 listens about as well as QWEN3.7 -- as in, not at all. It also gaslights you like Gemini about rules in the project, that it itself wrote, due to a misunderstanding, so you have to absolutely babysit it when it writes project guidelines, and review it at every step.

[jral @jral]

From the videos and articles that state prompting is different, and to just let it go do it's thing -- that's bad advice, at least on High. Because it's going to do what it wants, iterate in cycles, moving things around and declaring solved like it's ChatGPT with it's horn emoji.

[Conor Browne @brownecfm]

I know I've said this many times before on this platform, but the moment I learned anosmia was a symptom of Covid-19 was the moment I decided I was going to do everything possible to avoid getting it. I knew anosmia was an early prodromal symptom of Parkinson's Disease.

[Pixel FX @PixelFXco]

Morph 2K pre-orders are now open!

A $199 analog video scaler for retro gaming with composite, S-Video, SCART, component, and VGA support. Incredible video quality, CRT simulation, auto-sampling, Wi-Fi updates, and more.

To celebrate launch day, we're giving away one Morph 2K.

To enter:
✓ Follow @PixelFXco
✓ Like & repost this post

Winner selected June 20th.

https://t.co/xQqM20tdnt

[jral @jral]

@Dykeocletian @Y40IFRQTTING Search engines are basically useless, so you have to lean on them for anything 2022+. They’re also okay to ask about algorithmic things or a biology questions with citation. Beyond that though, these black boxes are dangerous if you don’t also read other material along side them.