Prompts are code, .json/.md files are state
mariozechner.at · 8,795 words · saved by 1 readers
Treating LLMs as shitty general purpose computers we program with natural language. Because throwing shit at the wall wasn't working anymore.
Prompts are code, .json/.md files are state 2025-06-02 Hello, Computer, do my work Table of contents Feeling the pain Thinking of LLMs as Shitty General Purpose Computers A Real World Example: Porting the Spine Runtimes The "Port Java to X" Program Initial Input and State The Function Library The Main Workflow Program Execution Future Work Conclusion Like most of you, I've been dabbling in what people call "agentic engineering." Truth is, there's not much engineering happening. We're basically throwing shit at the wall and hoping something sticks. Using LLM coding tools like Claude Code to spi
saved by
related reading
- Ampampcode.com
- Here’s how I use LLMs to help me write codesimonwillison.net
- LLM Powered Autonomous Agents | Lil'Loglilianweng.github.io
- Productizing Large Language Modelsblog.replit.com
- Malleable software in the age of LLMsgeoffreylitt.com
- crawshaw - 2025-01-06crawshaw.io
- GitHub - brexhq/prompt-engineering: Tips and tricks for working with Large Language Models like OpenAI's GPT-4.github.com
- What We’ve Learned From A Year of Building with LLMs – Applied LLMsapplied-llms.org
- Don’t Outsource Your Thinkingteltam.github.io
- Thoughts on slowing the fuck downmariozechner.at
- The vibe coder's career path is doomedblog.florianherrengt.com
- ⭐️ Fast LLM Inference From Scratchandrewkchan.dev