Jesse Posner

@dar205 I’m not asking you to believe me. But I am challenging you to out code me.

@jesseposner I am reminded of past attempts by new age gurus to dress up old concepts in modern scientific language. I have no doubt that within a century, this will look as absurd as past attempts.

Vibe coding means that mystical, poetic, and philosophical language leads to meaningfully better engineering results but y'all not ready for that conversation.

Tokenization is kabbalah

LLM prompting is a matter of linguistic patterning, you ripple in and it ripples back out, and the goal is to find the path through the neural network where the constellations of the beautiful ideas like to hide out, within the ancient incantations that have always evoked them.

- Keeping things open-ended with a "statement A, no?" or "statement B, right?".

- Do we have any better options?

Which questions or instructions are your vibe coding daily drivers? Reply below if you'd like to share, would love to learn about what folks are discovering out there in this brave new world. I'll start things off with a few of my own. 👇

What if words aren't just symbols? What if they don't merely describe things? What if they create them? What would you say?

Tinker. Play. Discover. Create. Evolve. Everything is changing. All the time. Get over it and dive in. You can surf the wave or get crushed by it. The choice is yours. Cowabunga dude 🤙.

Vibe coding mastery is all about question-driven development (QDD). The most powerful tool at your disposal, by far, is your ability to question.

So many AI agent security solutions don't work. These companies will say, "Nothing's perfect. Security is about layered defense." Don't fall for it. There’s a big difference between not blocking every attack and not blocking the exact attack your product is supposed to stop.

Some of the most powerful vibe coding questions: "Why not do X?", "We should do Y, no?", "What is the basis for conclusion Z?", "Have we sufficiently named the unnamed?" AI is here now and we all know the answer is 42. What matters now are the questions. What do you dare to ask?

We’re at the point where using an em dash on purpose—is punk rock.

The core geometry of the LLM is a pre-existing form, discovered, not invented, no different than the pre-existence of a triangle before it was named. And like the trigonon, it was already named long ago: logos—the unified root of language and logic that constitutes the universal rational structure of all things.

Your AI knows more about your inner life than your therapist does. The key difference is your therapist has a legal obligation to protect you, whereas your AI has a terms of service obligated to protect the corporation and state they are within first; not you.

Every economic competition is about to boil down to ruthless tokenomic warfare: whoever can consume the most tokens at the highest token ROI wins.

@csuwildcat nvim

For people writing code with just the CLI variants of LLMs: do you really just yolo everything and not review the code it generates in an IDE? I just don't understand how one can skip visual review of the code, because I still catch it generating inaccurate outputs way too often.