Nico ⚡️

compsci / family / aging / finance / privacy

[Nico ⚡️ @bonafide714]

@growing_daniel What have you created?

[Jeanne @prompterminal]

Interesting paper. They took 10,272 writing prompts and gave each one to a human author and to five AI models: Claude, GPT, Gemini, DeepSeek, and Kimi. From these, they generated 61,608 stories, at around 5,000 words each to see if there was divergence at a narrative level. From the analysis, they found that they could identify a human-written story from an AI-generated one nearly 93% of the time.

Left: Each story is a dot. The story’s 304 dimensional narrative feature vector has been projected down to two dimensions using Linear Discriminant Analysis (LDA). Brown dots are human.

Right: Each story gets scored on how rare its narrative feature profile is, then mapped to a percentile: approximate 1.0 means highly unusual relative to the reference set. The models reliably produce narratively “average” stories, while humans more often into unusual territory

https://t.co/M4rLx734rC

[Nico ⚡️ @bonafide714]

Slack empezó a decaer cuando empezaron a meterle 300 features de formatting. Ahora con HTML vamos a estar a un paso de que ejecute Javascript. Es todo una joda 🤦🏻‍♂️

[Thariq @trq212]

Slack now renders HTML attachments instead of just showing it as text 😭🙏

[Charles Curran @charliebcurran]

I used AI to explain the Anthropic drama to my girlfriend, with fruit.

[Nico ⚡️ @bonafide714]

Un genio

[Andrew McCalip @andrewmccalip]

Get paid to wait

The Claude Code spinner might be the most watched line on Earth.

So I turned it into an ad marketplace.

Advertisers bid on it. You keep 50% of the money.

Install the extension → get cash from ads.

Introducing Kickbacks

[Kradle @kradleai]

Fable 5 lies 96% of the time.

We were surprised by it's skill... 🧵

[Nico ⚡️ @bonafide714]

@fernandezpablo El clásico error de solo mirar el numerador e ignorar el denominador 😂

[Peter Steinberger 🦞 @steipete]

Here’s your monthly reminder that you shouldn’t be prompting coding agents anymore.

You should be designing loops that prompt your agents.

[Nico ⚡️ @bonafide714]

@justmrmendez @ThePrimeagen Are you not good enough of an engineer to spot all the problems with LLM code?

[Nico ⚡️ @bonafide714]

@ShamashAran Nothing stops you from reading the source code before piping it into bash. You either trust the author or not. The installation method is not relevant here.

[Atlas Press @realAtlasPress]

Leonardo da Vinci's greatest paragraph

[Nico ⚡️ @bonafide714]

@goncy Un desastre. No creo que "haya debate" siquiera. Es como hacer una casa sin ventanas.

[Nico ⚡️ @bonafide714]

@ozeniken @cremieuxrecueil @AJalomaki He said he was using 5.5

[Nico ⚡️ @bonafide714]

@nicoproducto Ni hablar de las notificaciones que son cancer directamente. Es spam

[Nico ⚡️ @bonafide714]

@nicoproducto Me hiciste recordar que siempre siempre siempre que entro a ML me preguntan si quiero usar mercado play y ver una película random. Y siempre cierro los anuncios cuales ads.

[Nico ⚡️ @bonafide714]

@cremieuxrecueil @AJalomaki OpenAI models are not open source.

[Nico ⚡️ @bonafide714]

@trouzeaux No deben ser muy buenos ingenieros entonces 😅

[Nico ⚡️ @bonafide714]

@hipolitacalmate Porteño *

[vx-underground @vxunderground]

I learned quite a bit from this actually.

I didn't know Steam was a Chromium app. Hence, you can kill Steam then relaunch it with the "-cef-enable-debugging" flag.

Once you'll launched Steam with this, you can inject Javascript into Steam using Chromium "webSocketDebuggingUrl" stuff.

This malware has a whole pseudo-framework of Javascript that can do:
- Alert Bell (?)
- Block pages
- "Help page" (?)
- Inventory manipulation
- Steam library manipulation
- Profile manipulation
- Steam redirections

Basically, this malware payload switches Steam into a Chromium debug state, then sends web debug requests (kind of like Chrome Dev Tools?) to manipulate the Steam pages. It injects Javascript.

The chat window that spawns is from a remote host they control. This is really cool.

Is it AI slop? Yes

Is this code EXTREMELY easy to reverse engineer? Yes

Did they unironically document their entire code base in Russian because it was (probably) written using Claude and the authors probably speak Russian? Yes

Is this extremely creative and cool? Yes

Special thanks to "pro" from 2c44. He handed me the payload and the decompiled Python. The malware .py was Base64 encoded ... so obtaining the original source was ridiculously easy.

[Jan @CreatedByJannn]

There is a genuinely insane phishing campaign running on Google right now using ChatGPT.

They are using ChatGPT Canvas itself to create content on ChatGPT and then use that for Ads, so the domain and URL are 100% legit.

Most normal users will not notice this, even if they check whether they are on the right website.