This week the headline is simple. If it works on us, it works for you. So before we sell branding to anyone, we are going to show you how we built our own. Branding is not a logo and a color. It is the set of rules that make a stranger trust you in three seconds. Today I will walk you through the whole system we run on at The Big T-M, and I will name every tool, so you can copy it. No mystery. Just the moves.

Start in a plain text file, not a design app. We open one markdown file and write three things: who we are, who we are not, and the words we never say. That is the foundation. We keep ours in the repo as brand-rails dot md. We even list banned words, like the hype words I am not allowed to use right now. When the rules live in text, a model can read them and stay on brand for you.

Next, we turn those words into numbers. Colors, fonts, and spacing become design tokens, which is just a small json file. Red is a hex code. The font is named once. The gap between things is one value. We feed that token file to every tool. Change one number, and the whole brand changes everywhere at once. That is how one person keeps a brand tight across hundreds of pieces.

Now we hand both files to Claude Code, running as a coding agent in the terminal. It reads the rules and the tokens and keeps them in mind for the entire session. So when we ask for a new post, it already knows the red, the font, and the words we banned. You are not re-explaining your brand a hundred times. The agent remembers it once and applies it every time.

For motion and video, we do not drag boxes in an app. We use Remotion, which lets you build video in React using your tokens. The same red, the same font, pulled straight from the json. The agent writes the composition, we press render, and out comes a clip that matches the brand by default, not by luck. Code is just a stricter designer that never forgets a rule.

Here is the part most agencies hide. We keep one source file that lists every post for the week. The agent reads that file and builds the social cut, the email, and the blog from the same brand. They rhyme because they come from one place. You write the idea once. The system dresses it for each room. That is consistency you can actually keep up with.

Before anything ships, we run a check. A second pass, a simulation, reads each piece and asks: does this match the rules file? Did it use a banned word? Is the red the right red? We built ours in python under an engine slash sim folder. It catches the off-brand piece before a human ever sees it. A brand holds because something is always watching it, even at three in the morning.

Then we publish it where you can see it. The whole studio runs in the open at office dot temerarii dot xyz. Every rule, every piece, every check. We did not do that to brag. We did it because a brand you can inspect is a brand you can trust. If we hid the machine, you would have no reason to believe it works.

Here is the honest proof. This whole calendar, every week of it, was built from one source file by the system I just described. No big team standing behind a curtain. You can go look right now. The brand you are watching is the brand we are teaching. That is the only proof we trust: the thing is live, and you can poke it.

So here is your next step. Do not copy our red. Copy the order. Write your rules in plain text. Turn them into tokens. Give both to a coding agent. Let it build and let a second pass check. Then put it somewhere people can see. See the whole thing running at office dot temerarii dot xyz, and build your own version of it.