A curated learning path for developers building generative AI and agent systems
When faced with a vast amount of AI material, this list selects only the core resources developers need to work directly with language models and agents. It is not a simple collection of links but a curated list verified weekly by an automated system based on technical depth and practicality. The resources are arranged to match a developer's actual workflow, from foundational theory to production deployment.
Selection criteria from a developer's perspective
This list is not a catalog of AI products. It includes only the resources necessary for software developers to understand, build, and deploy generative AI and agent systems. Every item must pass absolute quality standards of technical depth, practical value, evidence, and distinctiveness. The list follows the principle that an empty category is better than filling it with weak options. An evidence-based automated system independently reviews and verifies resources weekly to merge changes.
Structure connecting learning and building
The resources are divided into learning, system building, and agentic software engineering. The learning section includes books and courses covering everything from classical AI theory to the latest language model implementations. The building section covers guides and frameworks needed for actual application development, such as prompt engineering, retrieval-augmented generation, evaluation, and deployment observability. The agentic software engineering section introduces methods and related tools for improving software development itself using coding agents.
Practical patterns for agent systems
Specific patterns and protocols for building agent systems are emphasized. It includes agent building guides from Anthropic and OpenAI, as well as interoperability specifications like the Model Context Protocol and Agent2Agent Protocol. It also covers durable execution patterns for long-running agents, the design of human intervention points, and search and data processing tools, while broadly covering evaluation methodologies and observability tools that improve agent reliability.
Points to verify when selecting resources
This list is intentionally kept selective. When proposing a contribution, you must clearly explain the problem the developer is trying to solve, how it differs from existing items, and provide evidence of why it meets the quality standards. Resources must be up-to-date or have foundational, sustainable value. This list does not aim to promote specific products but focuses on improving developers' practical capabilities. The accuracy and maintenance status of the resources are verified periodically, making it suitable for checking the latest information.
