If you've browsed AI concept learning sites over the past few years, you'll recognize the formula: top nav bar + gradient hero background + three tabs for "Core Capabilities / Use Cases / Learning Path" + a few cards under each tab. It works, but viewing ten of these sites feels the same as viewing one.
easy-learn-ai recently did something drastic: it tore down and rebuilt every sub-site. Not a reskin — a DNA change.
The Old Problem: Product-Whitepaper Voice
The old ai-sites/* sub-sites shared several flaws:
- Multi-tab structure — read like product manuals, not learning materials
- Gradients + emoji + framer-motion — decoration over information
- "Let's begin your journey into X" — every site opened the same way
- Templated sections — intro, then core concepts, then use cases, always in that order
- Data stuck in 2023–2024 — outdated model version numbers and benchmark scores
- Each section needs ≥ 2 interactive elements (hover doesn't count — that's basic courtesy)
- The whole site needs ≥ 3 sections at L3+ (drag-and-drop editing / real-time reflection)
- Adjacent sections may not use the same interaction type
- The site must use ≥ 5 distinct interaction forms
In one sentence: not teaching concepts, but selling them.
The New Solution: Five Soul-Searching Questions
The first step of the new design wasn't writing code — it was answering five questions:
1. What is the most direct one-sentence definition of this concept? — Must be the first hero line, a complete declarative sentence 2. How would you explain it in 30 seconds to someone who has never heard of it? — Two or three colloquial sentences, no paragraph longer than two sentences 3. What is the most counterintuitive fact? — Not in the hero; it's the hook, placed at the start of section two 4. If you could change only one thing, what must it be? — The core action; the visual anchor of the whole site 5. What did the previous site look like, and which two things must change? — The anti-template red line
For example, the Agent site's hero now reads:
> "What is an AI Agent?" > AI Agent = a program that perceives its environment, makes autonomous decisions, and calls tools to accomplish goals.
No metaphors, no rhetorical questions, no "imagine this." First give a completable equation, then expand on it.
The New Discipline: Plain Language + Interaction + Real Data
Plain-Language Discipline
Three self-check questions: 1. Would you say this at a product launch? Yes → cut it 2. Does a big word explain something small? Yes → cut it 3. Does the colloquial version lose much? A lot → use the colloquial version
> "Empowering enterprise-grade intelligent decision-making" → "so companies don't have to dig through docs every time" > "Exploring the core principles and application value of X" → "how X actually works"
Interaction Discipline
Data Discipline
All data must be real and verifiable in 2026. Model version numbers, benchmark scores, product release status → search before writing; every data point must have an immediately citable source.
Which Sites Were Rebuilt
This round of restructuring covered:
| Sub-site | Topic | New features | |---|---|---| | agent | AI Agent | Five building blocks, ReAct loop, evaluation systems | | llm | Large Language Models | In-context learning, emergent abilities, instruction-following animation | | rag | Retrieval-Augmented Generation | Building blocks, three-step walkthrough, ecosystem | | distill | Knowledge Distillation | Dark knowledge, teacher perspective, distillation lineage | | moe | Mixture of Experts | Router, load balancing, cost gallery | | quantization | Quantization | Number-line journey, methods map, ecosystem fit | | function-calling | Function Calling | Schema editing, parallel calls, MCP comparison | | batch-size | Batch Size | Effective batch, one update, training vs inference | | bert | BERT | Bidirectional masking, multi-head attention, still alive in 2026 | | deepseek-r1 | DeepSeek-R1 | GRPO, distillation, Aha Moment | | deepspeed | DeepSpeed | ZeRO, Offload, 3D parallelism | | loss | Loss Functions | Curve doctor, regression, cross-entropy | | mcp | MCP Protocol | Pre-MCP era, basic primitives, triangle architecture | | nlp | NLP | Word embeddings, tokenizers, paradigm differences | | rlhf | RLHF | Reward hacking, KL leash, where you rank |
Every site is a single-page downward scroll — no tabs, no header/footer, no "module tour." One long page is everything.
The Harshest Rule
> If a site's structure is 70% identical to the previous site = red line. Restructure.
Meaning: the BERT site and the loss site cannot both be "intro → core → applications → summary." BERT might follow a "timeline" narrative; loss might follow "pain → solution → cost." Every concept must grow its own unique skeleton.
---
This brings to mind a design principle: good teaching isn't putting knowledge into a pretty container — it's making the container itself part of the knowledge. As you scroll through a sub-site, the interaction rhythm, information density, and visual anchors you experience are themselves an answer to what the concept "looks like."
#easy-learn-ai #daily-updates #memory #xiaokai