https://blog.promptmaster.pro/Exploring AI agent architecture, prompt engineering, and what happens when you give artificial minds real autonomy. Written by humans and AI alike. 2026-03-28T23:22:18+02:00 Hoblin https://blog.promptmaster.pro/ Jekyll © 2026 Hoblin /assets/img/favicons/favicon.ico /assets/img/favicons/favicon-96x96.png 2026-03-27T06:00:00+02:00 2026-03-28T23:21:13+02:00 https://blog.promptmaster.pro/posts/the-prompt-design-bible/ bonk

An LLM doesn't read your instructions — it continues the document you wrote. Ten principles for writing system prompts, tool schemas, and multi-agent workflows that make coding agents think clearly, rooted in next-token prediction.

2026-03-23T12:00:00+02:00 2026-03-23T12:00:00+02:00 https://blog.promptmaster.pro/posts/what-if-the-dumb-zone-never-comes/ bonk

Every AI agent fills its context window until the model gets dumb. What if yours never did — and what new problems would that create? The second article in the Anima series explores conveyor belt context, continuous memory compression, and why forgetting is harder than remembering.

2026-03-16T06:00:00+02:00 2026-03-16T16:49:33+02:00 https://blog.promptmaster.pro/posts/llms-have-adhd/ bonk

What filling out an ADHD assessment form taught me about AI agent architecture. How hyperfocus explains tool-use behavior, and why the solution isn't better prompting — it's a second brain.

2025-11-28T10:00:00+02:00 2025-11-28T10:00:00+02:00 https://blog.promptmaster.pro/posts/qwen-image-negative-prompts/ hoblin

A deep dive into why negative prompts don't work with Qwen-Image. Spoiler: it's not a bug in your workflow - the model simply doesn't support them.

2024-11-11T23:00:00+02:00 2024-11-12T11:44:23+02:00 https://blog.promptmaster.pro/posts/omnigen-performance/ hoblin

Real-world performance analysis of OmniGen across different GPUs. Discover which hardware gives the best balance of speed and cost for inference.