One Permission System for Three Coding Agents
Most people run coding agents on Full access because approval prompts wear them out. I rebuilt the permission system behind Claude Code, Codex and my own runtime around one goal: prompts you still read.
00 / Introduction
I build
software
I want
to exist.
Software developer
Open-source builder
Curious about how things work
Currently building Prokop and writing about AI agents, TypeScript, software ownership, open source, and shipping ambitious systems alone.
Daniel Bilek
Prague, Czech Republic
2026
01 / Selected
Featured writing
- 01
One Permission System for Three Coding Agents
Most people run coding agents on Full access because approval prompts wear them out. I rebuilt the permission system behind Claude Code, Codex and my own runtime around one goal: prompts you still read.
- 02
AI Won’t Make Us Smarter by Default
AI gives us access to more knowledge and capability than ever. That does not mean we will understand more, think better, or remain in control.
02 / Dispatches
Latest
Treating an LLM More Like a Human Produced Better Results
An agent's experience and a project's accumulated knowledge are not the same thing. Treating them as separate scopes changed how my agents work.
AI Won’t Make Us Smarter by Default
AI gives us access to more knowledge and capability than ever. That does not mean we will understand more, think better, or remain in control.
04 / Current work
Project
● Active development
Prokop
Open-source agentic coding workspace for persistent agents, parallel sessions, and inspectable context across projects.
Project details05 / Contact
Have something worth discussing?
Freelance development, collaboration, speaking, podcasts, technical conversations, or a concrete disagreement.
Get in touch