@hanakoxbt: KARPATHY JUST KILLED THE PROMPT ERA WITH A SINGLE DOCUMENT prompts are easy. loops are hard. and writing fifty prompts …
Summary
Andrej Karpathy releases a document describing an agent architecture that shifts from prompt-based interactions to a loop-based system where the model handles execution and bookkeeping while the human owns the spec and boundaries, effectively turning Claude into an autonomous agent.
View Cached Full Text
Cached at: 06/29/26, 06:24 AM
KARPATHY JUST KILLED THE PROMPT ERA WITH A SINGLE DOCUMENT
prompts are easy. loops are hard. and writing fifty prompts a day is the work nobody does twice.
he shifts the burden to the harness.
you define the contract once. the model writes, reviews, restarts, and reconciles. you keep judgment. it keeps the loop.
the throughline is the same in every rule: the human owns the spec and the boundary. the model owns the execution and the bookkeeping.
planner never touches code. generator never grades itself. state lives on disk, not in context.
9 rules. start with one feature, not ten. most people are still typing prompts. this turns Claude into an agent that finishes the job on its own.
here is the official document from Karpathy explaining the architecture
Similar Articles
@heyrohitai: Anthropic just showed a 27-minute workshop on how to actually write prompts for Claude. Taught by the people who built …
Anthropic held a free 27-minute workshop on prompt engineering for Claude, taught by its creators, offering valuable insights that are highly recommended for developers.
@Dipanshu_AI: Anthropic engineer: "at Anthropic, we don't write prompts anymore. we build loops" in 42 minutes, she shows how the Cla…
An Anthropic engineer explains that instead of writing prompts, they build loops, and demonstrates how the Claude team creates self-prompting loops in a free 42-minute video.
@RoundtableSpace: LOOP ENGINEERING IS REPLACING PROMPT ENGINEERING Repo: https://github.com/cobusgreyling/loop-engineering…
A new paradigm called loop engineering is introduced, shifting from manual prompt engineering to designing systems that autonomously prompt AI agents. The GitHub repository provides patterns, quickstarts, and npm packages for implementing loops with coding agents like Grok, Claude Code, and Cursor.
@polydao: THIS 11-PAGE CLAUDE LOOP ENGINEERING PLAYBOOK IS THE CHEAT CODE FOR SENIOR DEV PROMOTIONS typing prompts line by line w…
A playbook for designing self-prompting loops with Claude AI agents, covering discovery, handoff, verification, and persistence/scheduling for automated code development.
How I Think About Prompting AI Agents Across the Entire Prompt Hierarchy
The author shares principles for writing effective prompts for AI agents, emphasizing focusing on what truly matters, high-signal communication, actionable instructions, and using established phrasing.