code firstno hype
Agentic
Loops
Design patterns, system architecture, and production-grade implementations for software engineers building autonomous AI systems.
decides what to do
add result to context
↺
repeat until doneContext
System Instructions
Tools Definitions
User Message
Tool Result
LLM
Tools
web_search()
execute_code()
read_file()
tenets
Our Beliefs
Building AI agents is engineering, not magic.
LLMs are powerful, but on their own they're unpredictable and fragile. Building reliable agents takes real engineering — constraints, failure handling, testing, and clear system boundaries. We treat agents as software systems, not black boxes.
01. First principles over frameworks — Frameworks come and go. Context limits, statelessness, and hallucination don't. Master the constraints, not the hype.
02. Design for non-determinism — Traditional software gives you guarantees. LLMs don't. Architect for the uncertainty, don't pretend it away.
03. Master the trade-offs — Writing code has never been easier. Making the right architectural trade-off has never been harder. Autonomy vs control, flexibility vs predictability, speed vs safety — every decision demands deep understanding of the constraints underneath.
04. Learn by building — Reference implementations, battle-tested patterns, and first-principles thinking. No magic prompts. No black boxes. Just engineering.
TENETS.json
{
"belief": "engineering > hype",
"approach": "first_principle_thinking",
"determinism": false,
"accept_constraints": true,
"constraints": [
"context_limits",
"stateless_inference",
"hallucination",
"no_native_memory",
"prompt_injection"
],
"require": [
"context_engineering",
"tool_engineering",
"reusable_patterns",
"security_as_first_class_citizen",
"deliberate_trade_offs",
"traceability_and_auditability"
],
"reject": [
"prompt_and_pray",
"demo_driven_design",
"unverifiable_autonomy",
"trust_by_default"
]
}team
Who We Are
code
Featured Repos
We focus on internals, so engineers understand how agents work.
topics
What We Cover
A comprehensive curriculum for building production-ready AI agents
Loading course content...
Stay in the Loop
Subscribe for early access to repos, patterns, and engineering deep-dives before they go public.
Engineering-first. Production-grade. No magic.