deep-explainer
解釋文。概念推導、核心認識、應用邊界。
structure
The explainer archetype teaches a concept by starting with a concrete problem, building intuition through examples, then generalizing to the abstraction. It answers: what is this? How does it work? When would I use it?
- Concrete case study — real-world scenario that motivates the concept
- Where naive approaches fall short — why you can't solve it with obvious methods
- The core idea — the conceptual breakthrough
- What it actually looks like — worked example (code, diagram, or implementation detail)
- When you'd reach for it — application scope and boundaries
Works for: explaining algorithms, data structures, design patterns, academic concepts made practical.
required widgets
- Conceptual diagram — illustrating the core idea's structure
- Code block or worked example — showing implementation or pseudo-code
- Visual boundary marker — scope diagram or checklist showing when to use vs when to avoid
visual silhouette
archetype-check rules
- Exactly 5 H2s in order: "start with a concrete case", "where today's tools fall short", "the core idea", "what it actually looks like", "when you'd reach for it"
- `<p class="vg-deep-opener">` exists
- `<span class="vg-dropcap">` exists with exactly 1 character
- ≥1 `<pre><code>` block inside "what it actually looks like" section
- ≥2 inline `<svg>` elements (concept diagram + scope boundary)
- `<p class="vg-deep-closer">` with `<strong>Take-away</strong>`
- Universal contract: opener, dropcap, closer with strong tag
when to use
Explainer is the "teaching" archetype. Use it when your goal is to increase understanding of a concept, not just report news about it. The reader should be able to implement or apply the concept after reading, not just know it exists.
Posts using this archetype
- 06.17 HTTP QUERY 方法(RFC 10008)
- 06.16 TimescaleDB 怎麼把時序壓到剩零頭——columnar 壓縮從零講起
- 06.15 私有推論還不夠私有——Siri 新架構的威脅模型缺口
- 06.14 普查局想拿掉差異隱私的雜訊——為什麼統計學家急著反對
- 06.12 invisible dependency,從零講起——你的模型到底建在誰之上
- 06.11 DiffusionGemma——把文字生成從自迴歸換成擴散去噪
- 06.10 Postgres 終於要做 query hint——pg_plan_advice 怎麼回應十五年的六條反對
- 06.08 加 queue 救不了過載——為什麼緩衝只是把崩潰往後延,以及該做什麼
- 06.06 durable execution 搬進 PostgreSQL——pg_durable 用一張 schema 接住 crash
- 06.05 KV cache 為什麼能壓到 4-bit/2-bit 還不掉準確度——從 variance normalization 講起
- 06.04 後量子憑證為什麼不直接簽——Merkle Tree Certificate 怎麼把握手壓回比今天還小
- 05.28 Atom 用完,VM 就掛——拆解 BEAM 那張全域 atom 表
- 05.26 把 LLM 訓練當成有雜訊的通道——Shannon Scaling Law 怎麼解釋過訓練與量化退化
- 05.25 InnoDB 的 next-key lock,從零講起——REPEATABLE READ 怎麼用 gap lock 擋掉 phantom read
- 05.23 把 LLM eval 擺在 A/B 前面——funnel vs fork、calibration、與那 42% 還是被 rollback 的實驗
- 05.20 Quint,從零講起——TLA+ 風格 formal model 怎麼幫 Turso 找出 SQLite 十個 bug
- 05.18 agent 的上下文管理,從零講起——Slack 為什麼不傳訊息歷史
- 05.17 std.Io,從零講起——Zig 0.16 怎麼把非同步從語言降級成庫的選擇