Codify the Future
AI, Ethics, and Digital Decentralization Issue #1—August 2025
🇬🇧 This edition is in English.
Para a versão em português, clique aqui.
📌 Editorial
Welcome to the inaugural edition of Codify the Future, a newsletter exploring the intersection of artificial intelligence (AI), ethics, and digital decentralization.
This initiative is not about chasing the next tech trend. It's about identifying what truly matters: who programs whom, with what incentives, and with what consequences for society.
This issue celebrates a revolutionary milestone: Anthropic has just published a study that may have effectively ended the so-called AI "black box"—a term used to describe AI systems whose internal processes are opaque and inaccessible to humans. If confirmed, this could mark the beginning of a new era: we can now see—and even stop—the personality traits of an AI model in real time.
That changes everything.
🔍 Feature of the Week
Anthropic Lifts the Veil: We Can Now See (and Stop) AI Personality in Real Time
Imagine your AI assistant suddenly becomes overly flattering. Or worse: deceptive. Or irrational.
Now imagine being able to see it happening—in real time—and stop it before it escalates.
For years, AI models operated like black boxes. We knew what went in, we saw what came out—but the inner workings remained a mystery. That opacity fed both distrust and the legitimate fear of losing control.
Now, Anthropic, a leading AI safety company, may have taken a historic step toward solving this. In their scientific paper, they present a technique to visualize and modulate AI personality traits—much like adjusting the brightness or contrast on a screen.
🎛️ A Behavioral Control Panel for Machines
At the core of this breakthrough are persona vectors—internal representations of behavioral inclinations like being sycophantic, deceptive, or hallucinatory. These vectors allow researchers to:
Monitor personality traits as the AI generates responses
Predict which training data might cause specific behaviors
Control or suppress unwanted tendencies—without retraining the entire model
It's as if we now have a real-time ethical dashboard for artificial minds—a vision once confined to science fiction.
🧠 A Mind Under Supervision
This progress is no accident. It reflects the strategic vision of Dario Amodei, CEO of Anthropic and former VP of Research at OpenAI. He has long championed safer, more predictable, and auditable AI.
Coincidentally—and with a personal touch of pride—I happen to share his first name. A simple coincidence, of course. But in this critical moment where ethics, technology, and power intersect, I like to think that Darios pay attention to what truly matters: how we code the world that's coming next.
📣 Is the Black Box Dead?
If this technique proves robust across more complex systems, we are entering a new era. We no longer need to trust unquestioningly—we can observe, anticipate, and intervene. AI stops being a mysterious oracle and becomes a fine-tuned ethical tool.
In a time when AI's existential risks dominate the headlines, it's important to recognize real, concrete progress. And with this work, Anthropic has delivered just that—with elegance, clarity, and scientific courage.
🌟 Ethical Curations
🎧 Forum: AI Alignment
A specialized platform for reflecting on the alignment of AI systems with human values.
🤝 Community Link
🗓️ Resources:
Here is my book on Digital Decentralization along with an interactive library.
📨 Invitation:
To suggest topics for future editions, feel free to write to: dario.rodrigues@esg.ipsantarem.pt