A free, public training platform for learning AI — beginner to master. Curated external resources where they exist, original content where they don't, with the working discipline that turns study into capability.
Mastery is what happens during the years of work that come after a curriculum ends. This platform gives you the curriculum part — verified resources, working artifacts at each stage, paper reproductions, a contribution path. The rest is yours.
Anyone serious about learning AI:
- Total beginners — no prior ML or programming background assumed.
- Software engineers adding ML / AI to existing skills.
- Career-switchers moving into AI from another technical or analytical field.
- Mid-career engineers looking to push toward original work and frontier engagement.
The path scales:
- Foundations (beginner) — Python, math intuition, first ML, first DL.
- Foundations (intermediate) — math, classical ML, deep learning, transformers, RL, systems, interpretability.
- Specialization (advanced) — pick one of: LLM engineering · CV / multimodal · MLOps · AI safety.
- Electives — software-engineering hygiene, paper reproduction, OSS contribution, domain pairing, compute, community.
- Frontier — independent work, original contributions.
- Anchored on the source guide. Curriculum starts from The Complete Free AI & ML Guide (Edition I, April 2026) — a meta-research document that lists canonical free resources and corrects outdated ones.
- Verified, not exhaustive. Every resource added is vetted: free or substantially free, currently maintained or flagged as frozen, URL verified, pedagogy sound. Standards are in MAINTAINING.md.
- Curated now, original later. As the maintainer masters each area, original content (labs, exercises, writeups) gets layered on top of the curated path under a clear license — see below.
If you're new to programming and AI: part-2-beginner-ramp/ — Python, math intuition, first ML, first DL. ~3–6 months at a steady pace.
If you have a software background and basic ML literacy: part-3-foundations/ — math, core ML, deep learning, transformers, RL, systems, interpretability.
If you're past foundations and ready to specialize: part-4-specialization/ — pick one of four tracks.
For the working habits that turn study into expertise: part-5-electives/ — software engineering, paper reading, paper reproduction, OSS contribution, domain pairing, compute, community.
(Folder structure currently mirrors the source guide. A level-numbered scheme — 10-foundations-beginner/, 20-foundations-intermediate/, etc. — is on docs/roadmap.md.)
Read LEARNING.md. Short version: depth beats breadth, finish what you start, ship one artifact per stage, read papers from current venues, reproduce one per quarter, contribute one merged OSS PR per stage.
See CONTRIBUTING.md. Priorities: new resources that fill real gaps, broken-link fixes, voice cleanups, and (once core curriculum is stable) translations and original content.
When opening an issue, GitHub will offer three structured templates: resource proposal, broken link, voice cleanup. Pick the one that fits — they channel contributions into the right shape and reduce review friction. Pull requests use a checklist embedded in .github/pull_request_template.md covering MAINTAINING.md verification standards.
Community standards: CODE_OF_CONDUCT.md. Sensitive issues (PII, accidental commits, vulnerability-style problems): SECURITY.md.
- LEARNING.md — study advice for users of the platform.
- MAINTAINING.md — author / curator standards (verification, voice, no fabrication).
- CONTRIBUTING.md — how to propose resources, fixes, or original content.
- CODE_OF_CONDUCT.md — community standards (Contributor Covenant).
- SECURITY.md — vulnerability and PII reporting policy.
- docs/roadmap.md — public roadmap for what's next.
- docs/books.md — free books referenced across the curriculum, in one place.
- docs/domain-pairing.md — method + free public data sources for ML-plus-domain work.
- docs/conferences.md — free tutorial / workshop recordings from major AI conferences.
- docs/catalog.md — single quick-reference index of every resource with type / status / location.
- docs/faq.md — common questions about using the platform.
- docs/anti-patterns.md — failure modes in self-directed AI learning, with diagnostics and fixes.
- docs/stuck-diagnosis.md — when stuck mid-session, name the stuck-type and apply the matching fix.
- docs/track-selection.md — 21-question self-assessment + signal-vs-noise per track for picking a Part IV specialization.
- docs/paper-reading.md — Keshav's three-pass method adapted to deep-learning papers, with a worked example.
- docs/carbon-aware-ml.md — measuring and minimizing inference carbon footprint; tools (CodeCarbon, SEAL, LLMCarbon) and engineering patterns.
- docs/vibecoding.md — vibecoding vs spec-driven development; when to use which working with AI coding assistants.
- docs/reading-rubric.md — six-question framework for evaluating any course / book / paper before committing weeks.
- docs/prerequisite-map.md — concept dependency graph: what unlocks what across math, ML, deep learning, and applied AI.
- docs/consumer-agent-platforms.md — OpenClaw / ClawHub / ClawMoat ecosystem and where local-first agents fit in Track 1 deployment thinking.
- docs/prerequisite-stress-tests.md — concrete diagnostics per Layer of the prereq map; companion to prerequisite-map.
- docs/time-budget-calibration.md — Tetlock-style calibration training applied to self-directed learning; planned vs actual time journal protocol.
- my-path/ — the maintainer's personal trail through the platform; not part of the public curriculum.
- projects/ — artifact catalog.
- Curriculum: part-2-beginner-ramp/, part-3-foundations/, part-4-specialization/, part-5-electives/.
Anchor curriculum: The Complete Free AI & ML Guide — Edition I, April 2026. The PDF is not redistributed here; resources are referenced by name in each part's README, with verified URLs where the source guide spelled them out.
This platform is curriculum + structure + verified resources, not yet an example of someone using them well.
- The maintainer has not yet completed the curriculum themselves.
my-path/progress.mdandprojects/are scaffolds; they will fill as the maintainer ships work, not before. - Original synthesis content (anti-patterns, stuck-diagnosis, track-selection, paper-reading, prerequisite-map, reading-rubric, carbon-aware-ml, vibecoding, consumer-agent-platforms) is rigorous synthesis from public knowledge. It does not yet reflect the maintainer's work-from-experience. The framing is sound; the "this person did the work" signal is currently absent.
- The platform's strongest implicit claim — "the working discipline that turns study into capability" — is true of the discipline. The demonstration of the claim through the maintainer's own shipped artifacts is the gap to be earned over time.
- Zero real users yet. Structural decisions (subfolder layout, doc organization, voice) reflect maintainer + AI-assisted intuition about learner needs, untested by actual learners.
If you're evaluating whether this platform is for you: the curation discipline is real, the structure is sound, the frame is honest about its limits. The demonstration of mastery is something to be earned over time, not claimed prematurely.
Active build-out, April 2026.
- Curriculum scaffold mirrors the source guide's structure (Part II beginner ramp through Part V electives) with per-step and per-track subfolders.
- Verified resources — 23+ individually URL-checked in-session against MAINTAINING.md standards. A URL re-verification pass caught and corrected six real source-guide claims (Goodfellow PDF restriction, UDL URL path, Foundations-of-LLMs license, SLP date, Math-Foundations-RL repo path, Inspect AI URL).
- Original synthesis content — anti-patterns, stuck diagnosis, track selection, paper reading, prerequisite map, reading rubric, carbon-aware ML engineering, vibecoding, consumer agent platforms.
- CI — every push or PR runs markdown-link-check across the docs to catch URL rot automatically.
- Governance — public-repo standard set: code of conduct, security policy, issue templates, PR template.
Treat current curriculum as a working starting point. See docs/roadmap.md for what's gated for later, and docs/catalog.md for the full resource inventory with verification status per entry.