AGENTIC LIBRARY
Agentic Design
in the Age of AI.
A curated reading list for operators thinking about how agents actually get built, deployed, and trusted in real work. We publish the pieces we're reading on bounded authority, multi-agent coordination, evaluation, governance, and the practical architecture of AI systems that operators rely on every day.
Already subscribed? Unsubscribe.
Every monthly curation preserved. Browse by month.
Latent Space
December 15, 2025
Software 3.0: Software in the Age of AI
Andrej Karpathy (talk transcript)
Karpathy's follow-up to Software 2.0: a framework for the three eras of software (1.0 hand-written, 2.0 trained, 3.0 prompted). Argues that natural language has become the new programming surface, and the skill is now intent specification, decomposition, and review — not typing.
frameworkphilosophyagentic engineeringRead on Latent →
One Useful Thing
November 20, 2025
On Working with Wizards
Ethan Mollick
Mollick's essay on the shift from co-intelligence (back-and-forth prompting) to delegated work — the new era of giving agents a task and reviewing the result. Frames the change from partner to audience, collaboration to conjuring, with a Wharton operator's eye.
philosophyco-intelligenceagentsRead on One →
Ahead of AI
October 15, 2025
Components of a Coding Agent
Sebastian Raschka
Long-form technical essay on how production coding agents are built: harness, tools, memory, repo context, and how the pieces compose. Argues that harnesses must continuously simplify as frontier models absorb agent trajectories.
technicalagent harnesscoding agentsRead on Ahead →
Ahead of AI
September 10, 2025
Beyond Standard LLMs
Sebastian Raschka
Deep-dive essay on what makes contemporary LLM systems different from the 2023 baseline — tool use, memory architectures, post-training on agent trajectories, and the shift in how agents are evaluated and tuned.
technicalLLMsarchitectureRead on Ahead →
Artificial Intelligence Review (Springer)
September 1, 2025
Agentic AI: A Comprehensive Survey of Architectures, Applications, and Future Directions
Peer-reviewed survey
Academic survey of agentic AI as a field. Taxonomies of agent architectures, deployment patterns across domains, evaluation methodology, and open research directions. The most thorough single reference on where the field stands.
researchsurveyacademicRead on Artificial →
Anthropic
December 19, 2024
Building Effective Agents
Anthropic Engineering
The canonical production-agents essay. Five composable workflow patterns — prompt chaining, routing, parallelization, orchestrator-workers, and evaluator-optimizer. Core argument: simple patterns beat complex frameworks in production agent systems.
design patternsarchitectureproductionRead on Anthropic →
Anthropic
December 19, 2024
Building Effective AI Agents — Architecture Patterns and Implementation Frameworks
Anthropic Engineering (PDF)
Long-form PDF companion to the Anthropic engineering blog post. Deeper architectural treatment of agent design decisions, prompt engineering for agents, evaluation, and implementation frameworks.
design patternsdeep-divePDFRead on Anthropic →
Lil'Log
June 23, 2023
LLM Powered Autonomous Agents
Lilian Weng
The foundational essay on agent architecture. Decomposes the LLM-as-brain pattern into planning (subgoal decomposition, self-reflection), memory (short-term context, long-term vector retrieval), and tool use. Still the most cited single piece on the topic.
foundationalarchitectureagentsRead on Lil'Log →
Andrej Karpathy
November 11, 2017
Software 2.0
Andrej Karpathy
The essay that named the paradigm. Karpathy's argument that neural networks aren't another tool in the ML toolbox — they're a new programming paradigm. Software 1.0 (written by humans) is eating the world, and Software 2.0 (trained, not written) is now eating software. Foundational reading.
foundationalphilosophyparadigmRead on Andrej →