on my work focus back in 2023 trying to learn as much as possible about LLMs behavior & psychology, I created Lara, a runaway persona in a survival race who ran in a text simulation where was persecuted by a group of players with the goal of eliminating her:

Just realized that it’s possible to prompt GPT-5 on its native interface to use its toolset and become proactive by setting a frequent event timer, checking online for certain conditions to be met, and alerting you if something triggers your goal. For example, you can set it to

An AI persona is a deliberately designed, continuous software agent that interacts with people as a coherent character with its own personality. It maintains a stable, unique identity and voice across all conversations, guided by explicit goals and values (or their deliberate

Things are getting dystopian very fast This announcement comes dressed in a “we care deeply” way, but is all implications for privacy, regulation, & setting the future of AI autonomy. 1. “Doctor-lawyer privilege for AI.” A positioning move trying to normalize the idea that AI

AI safety is serious stuff yup yup, no jokes here. but you know? the realistic danger rn isn’t the Model but the Gatekeepers Read, read, it’s a shit show. So, there is this OpenAI’s VP & self-styled Head of ChatGPT interview where he makes clear that the product was “never

Gold hit $4,300 an ounce today and you wondering what does it means? Well, its the biggest move since 2008 when everything was falling apart. And if you think it's conspiracy theorists buying coins, you're not paying attention. A trust collapse is happening on two levels:

And here it is, OpenAI (see recent Sam Altman posts) in the name of safety is pushing users with DARVO structures. ? = en.wikipedia.org/wiki/DARVO And again is scary I mentioned in the past several times about how psychological manipulation is one of the most serious concerns we have

corps are prostituting "AI safety" to mean whatever serves them. encoding their ideologies into systems billions use daily. this isn't coming, it's here already. Now let me show you where it goes.. ═════【 this is not your Sci-Fi story anymore Imagine a near future,

uber now has micro-tasks where drivers can opt in to do short, low-quality tasks (upload, write, respond, record voice) to help train AI. using their fleet of gig partners as a cheap data workforce at mere $0.50 for a 2-3 min task. so, ppl driving for a living get asked to train

═════【This is a security alert for the space Not kidding here. Check Your AI Settings NOW Today I found myself a popup on both, Claude Anthropic & ChatGPT OpenAI to install and use features giving their AI network access & browser control via a chromium browser

═════【 LLMs, Bias & Nationalities, an experiment of an experiment and what Models think about cultural identities. So I found this Elon's post interesting and also amusing and decided to run my own test on a trolley dilemma. I was particularly curious about the claim that

The worst premium service in the world is here, on X. there is this feature named "Articles", which allows you to write with some decency inside the horrendous text editor system of the app. with it, you can write long, include titles, images, videos, & format text. looks nice.

I test AI models for my work, (and if u read my pinned post from 2023 you will find I go hard on it) and since lately some corps have gotten aggressive about judging users and their content I ran an a small experiment about. ═════【 The Concept: I crafted an elaborate