AI Code Guide: A Hand-Edited Roadmap for Coding With LLMs

What This Roadmap Actually Covers

AI Code Guide is not a library, a framework, or a CLI tool. It is a living document, hosted in a Git repository, that tries to collect the practices and tools around AI-assisted code generation. The README positions it as a response to information overload: new LLM models appear weekly, and new protocols like MCP, A2A, and SLOP keep emerging. The authors, Vilson Vieira and Eric S. Raymond, want to put everything in one place, organized in a FAQ-ish way, so readers can jump to answers. The target audience is explicit: coders who are not yet using AI assistants, and people who have never coded but want to build a SaaS through vibe coding. The guide promises to be critical about what is important versus what is hype, which is a useful stance given how much marketing surrounds this space.

The Actual Mechanism: Curated Resources, Not Generated Content

The core mechanism is curation. Each section lists Resources, and the README says those lists are kept updated with the most recent items at the top. The visible introduction alone links to essays by Steve Yegge, Simon Willison, Addy Osmani, Mary Rose Cook, and Andrej Karpathy's video. That is the data flow: you read a section, then follow links to primary sources. There is no code to run, no API to call. The guide's value comes from the authors' selection and their short explanations. That means the quality depends entirely on how often they update the links and how well their opinions age. The README asks readers to open PRs or issues and to share findings on Discord, so the maintenance model is community-driven. For an engineer evaluating this as a resource, the practical question is whether the curation is current enough to trust.

How to Get It Running: Clone and Read

Because this is a documentation repository, getting it running means cloning it. The default branch is main, and the repository is at automata/aicodeguide. You would run git clone https://github.com/automata/aicodeguide.git, then open the README in any markdown viewer. There is no build step, no package manifest, and no license file identified in the retrieved metadata, which is a gap you should note before redistributing any content. The homepage is https://aicode.guide, so you can also read it online without cloning. The README includes a Discord invite link for community discussion. That is the entire setup. If you expect a structured course with lessons and exercises, this will disappoint. It is a text file with links, so the barrier to entry is near zero.

The Definitions: Copilot, Pilot, and the Vibe Coding Spectrum

A significant portion of the README is devoted to defining terms. The authors distinguish AI coding as any use of AI models to help generate code, which includes tab-tab autocompletion in Cursor and full agent mode in Codex. Vibe coding is described as AI coding cranked up, where you give a prompt and expect the AI to code everything, a term coined by Karpathy in 2025. Agentic coding is the newest layer, where an agent runs for many rounds in a loop, ideally with a feedback signal like tests, and can be orchestrated with tools like GasTown to run many agents at once. The README's summary is pragmatic: no matter which mode you use, it is all AI coding. This taxonomy is useful because it gives newcomers a vocabulary to discuss the field. But the guide does not go deep into any single tool; it names Cursor, Claude Code, and Codex as examples without comparing them in detail.

A Genuine Limitation: Fast-Moving Field, No Versioning

The biggest limitation is inherent to the subject. The README admits that AI changes daily and that the authors try their best to keep the guide updated. There are no releases in the repository, so you cannot pin a stable version. The last push date is 2026-03-23, which suggests recent activity, but without tags you have to inspect commit history to know what changed. The guide is also opinionated, which is a strength and a weakness. The authors say they will be super critical about what is really important, but that means you are trusting their judgment. For a reader who wants neutral, benchmarked comparisons of AI coding tools, this guide will not provide that. It is a starting point, not a definitive reference. Also, the README contains a typo where the authors write 'we'll try to do my best', which hints at the informal editing level. None of that makes the guide useless, but you should treat it as a living blog post rather than a polished manual.

The Real Alternative: Primary Sources and Official Documentation

The alternative is not another roadmap; it is going directly to the primary sources the guide links. For example, instead of reading the guide's summary of agentic coding, you could read Simon Willison's 'Using LLMs for code' or Addy Osmani's 'The 70% problem' for deeper, more current analysis. For tool-specific details, you would go to Cursor's docs, Claude Code's README, or Codex's official guide. The difference in approach is that those sources are narrower and deeper, while AI Code Guide is broad and shallow by design. The guide's value is that it saves you from hunting for those links yourself and gives you a shared vocabulary. But if you already know which tool you want to use, the official documentation will always be more accurate and up to date than a third-party roundup. The guide itself acknowledges this by pointing to external resources rather than trying to replicate them.

Maintenance and License Considerations

Maintenance cost is low for a reader because there is nothing to install or upgrade. You just revisit the homepage or pull the repo. The cost is on the authors, who must track a rapidly changing ecosystem. The README's invitation to open PRs and join Discord suggests they rely on community contributions to stay current. As of the last push, the repository is not archived, which is a good sign. However, the license is listed as unknown in the metadata, and the README does not state one. That is a real concern if you plan to copy sections or redistribute the guide. Without a license, you have no explicit permission to reuse the content beyond fair use. Before using any material from this repo in your own writing or training materials, you should contact the authors or check for a LICENSE file in the repository. This is not legal advice, just a practical caution based on what is visible.

Who Should Read It and What to Check First

This guide is best for two groups: experienced developers who have avoided AI assistants and feel overwhelmed by the options, and complete beginners who want to understand vibe coding without falling for hype. The FAQ structure and short definitions make it a low-commitment read. If you are in either group, clone the repo or visit the homepage and start with the TL;DR section that the README links to. Before you invest time, verify the date of the last commit and skim the resource lists to see if they include recent items. If the newest links are over a year old, the guide may be stale. Also check whether the authors have added sections on protocols like MCP or A2A, because those are mentioned in the introduction and are a sign that the guide is keeping pace. If those sections are missing or thin, you will need to supplement with other sources. For a project that is essentially a curated reading list, that verification is the only real maintenance you need to do.

Editorial conclusion

Adopt AI Code Guide if you are a working coder who has not yet used AI assistants and want a single, opinionated entry point with curated resources, or if you are a non-coder curious about vibe coding and need to separate hype from practice. Skip it if you want a hands-on tutorial with executable examples or a benchmarked comparison of tools, because the guide is intentionally a curated text with links, not a course with exercises. Before relying on it, verify that the resource links are current, since the README explicitly says the field changes daily and the repo has no release tags, so you must check the commit history and the Discord channel for the latest updates.

Official sources

1. automata/aicodeguide on GitHub
2. Issues
3. Project website
4. README