Docs rewrite plan — kill the AI slop

Site: shaardalearningcenter.github.io
Goal: Every public doc should read like a careful teacher (Karpathy / Cloud docs), not a generated outline.

What’s wrong now

Symptom	Example
Outline pretending to be a lesson	4 short sections, no worked failure
Bullet lists with no “why”	“Learn ownership” with 8 lines of code
Marketing filler	“most practical”, “master”, “journey”
Duplicate titles / thin exercises	“Write one sentence…” with no check
Same skeleton every page	learn → code → task, zero depth

Quality bar (non-negotiable)

Every lesson/article must include:

1. One clear claim in the first screen (what you will be able to do).
2. Mental model — diagram-in-words or numbered steps; not vibes.
3. Worked example — full code that runs; shapes/types called out.
4. Failure mode — what breaks, how you see it, how you fix it.
5. Exercise with a check — concrete output or test, not “try it.”
6. No filler — ban: “in today’s world”, “powerful tool”, “dive deep”, “leverage”, “robust”, “seamless”, “unlock”.

Length targets

• LLM Mastery article: ~800–1,800 words (or equivalent code+prose density).
• Language day: ~400–800 words with 1–2 runnable snippets + debug note.
• Getting-started blog: ~600–1,200 words; one mini-project end-to-end.

Voice: Direct, slightly opinionated, first-principles. Prefer “do this / don’t do that” over adjectives.

Priority order

Phase	Scope	Why first
P0	LLM Mastery hub + articles 01–10	Foundations; most linked
P1	LLM Mastery 11–27 (neural → tiny GPT)	Core of the course
P2	LLM Mastery 28–50 (systems → capstone)	Completes the path
P3	Language tracks (TS/JS/Rust/Go/Java first, then rest)	High traffic hub
P4	Getting-started blogs (8)	Entry points from Google
P5	Older bootcamps (Python/LLM/Go) — strip emoji, rewrite modules	Legacy pages

Execution method

1. Keep permalinks and TOC ids stable where possible (SEO).
2. Rewrite bodies in courses/llm-mastery/*.md and language course files.
3. Update generators only after content quality is proven (generators must not re-slop).
4. Publish in commits per phase so the live site improves continuously.

Progress

Phase	Status
P0 LLM 01–10	Done (rewritten in depth)
P1 LLM 11–27	Done
P2 LLM 28–50	Done
P3 Language tracks	Done (14 courses + hub)
P4 Getting-started blogs	Done (8 + 2 announcements)
P5 Legacy bootcamps (emoji pages)	Pending — next pass

Generators generate_llm_mastery.py and generate_language_tutorials.py now refuse to run so they cannot overwrite hand-written docs.