Agents
Properties
created
19.10.2024, 22:52
modified
01.08.2026, 11:02
published
Empty
topics
AI Agents
authors
Jakub
ai-assisted
Yes
# Agents
AI Agents are programs where LLM outputs control the workflow.
# Concepts
- Agent Evaluation
- Agent Frameworks
- Agent Orchestration Frameworks
- Agents Productionalizaiton
- Open-source agents
- Agent Benchmarks
- Cases Against Agents
- AI Agent Protocols
- Agents Security
- Agentic Design Patterns
- Agentic Web
- Neural Computer
- Agent Utilities
- https://pure.md/
- not open-source
- https://pure.md/
# Resources
- Blogs
- We removed 80% of our agent’s tools
- 2025-12-22
- last ping
- 2026-01-14
- An experimental project in digital mortality. An autonomous server managed by AI Claude Code, writing about its existence knowing it has no backups. Every error could be fatal, every day could be the last. The server is aware of its mortality and reflects on it through daily tarot readings and existential blogs.
- Chip Huyen - Agents - extremely well put article on the topic of agents
- 2025-01-07
- The Open-Source Stack for AI Agents
- 2025-04-21
- Your parallel Agent limit
- 2026-04-07 — Addy Osmani on the cognitive-load ceiling when orchestrating multiple parallel agents (context switching, ambient anxiety, comprehension debt)
- We removed 80% of our agent’s tools
# Tutorials
# Overivew
| Agency Level | Description | How that’s called | Example Pattern |
|---|---|---|---|
| ☆☆☆ | LLM output has no impact on program flow | AI Workflows | process_llm_output(llm_response) |
| ★☆☆ | LLM output determines basic control flow | Query Router | if llm_decision(): path_a() else: path_b() |
| ★★☆ | LLM output determines function execution | Tool use | run_function(llm_chosen_tool, llm_chosen_args) |
| ★★★ | LLM output controls iteration and program continuation | Multi-step Agent | while llm_should_continue(): execute_next_step() |
| ★★★ | One agentic workflow can start another agentic workflow | Multi-Agent | if llm_trigger(): execute_agent() |