I’ll address that briefly here. I only have about four years of experience in a traditional corporate setting, so I wouldn’t call myself an expert at navigating political dynamics within a hierarchical structure (there’s still plenty for me to learn).

Most of my career, to this day, has been in consulting. So, when I was brought into “the room,” it was usually because I was the external person called in to clean up the mess. That meant I already had some weight behind me, and the dynamics were quite different from being part of the internal chain of command.

More often than not, the person who brought me in was the CEO, a board member, or a VP. I reported directly to them, which meant I carried their backing, and their weight with me. Being external to the team also gives you a bit of a cultural wildcard. You know you're only there for a limited time, so you're more willing to burn the bridges that need burning and cut through the noise to get things done.

To answer our reader’s question about how I “stood up to figures of authority,” the truth is actually pretty simple: I treated them like anyone else. I listened, kept asking “why” until they either gave up or opened up, and once we got to the root of their concerns, we worked together to come up with a plan to address them. It’s kind of like therapy, but it’s not.

How do you replicate this in a corporate setting, especially in a remote environment? That’s the million-dollar question. It takes time, patience, and a lot of intentional effort. Without face-to-face interaction, building trust becomes even trickier. You have to move beyond just saying the right things. It’s about doing the work, showing up, staying subtly visible, and staying consistent, often without any immediate feedback.

As I’ve said at the beginning, I don’t consider myself an expert in this specific environment, but I’d say the core principles still apply. That said, doing it remotely demands twice the patience, twice the energy, and a whole lot more consistency. And yeah, every now and then, there’s that familiar gut-check: “Am I the only one who actually gives a damn about this?”

In his last post, Armin writes:

As far as the core model is concerned, there’s no magical distinction between “user text” and “assistant text”—everything is just tokens. The only difference comes from special tokens and formatting that encode roles (system, user, assistant, tool), injected into the stream via the prompt template. You can look at the system prompt templates on Ollama for the different models to get an idea.

This is an accurate description of what I think a model is.