今日固定时间表
只有三个学习窗口：08:50-10:55、14:30-16:15、20:30-21:15。section links 只放在本日页面。
| 08:50-10:55 | System design 读 Ad Click Aggregator Potential Deep Dives：10k clicks/sec, no click loss, abuse/dedupe, low-latency metrics；做 hot shard split 方案。 |
|---|---|
| 14:30-16:15 | NeetCode Linked List 三道连续题位，每题 30m solve + 5m pattern/invariant/bug/complexity evidence。 |
| 20:30-21:15 | Recall / Staff Q&A / Mock Frontier radar gate：无一手 AI source 改变本周 contract，记录 intentional omission；不加泛 RAG/agent 阅读。 |
算法模块
连续 NeetCode Linked List tag；每天恰好 3 个 slot。
| # | Problem / mode | Pattern | Invariant | Bug risk | Time | Space |
|---|---|---|---|---|---|---|
| 10 | Merge K Sorted Linked Lists NEW · 30m solve + 5m evidence | heap by node value | heap contains next smallest candidate per list | pushing null or losing node.next | O(N log k) | O(k) |
| 11 | Reverse Nodes In K Group NEW · 30m solve + 5m evidence | bounded in-place reversal | only complete groups of k are reversed | reversing short tail | O(n) | O(1) |
| 12 | Swap Nodes in Pairs NEW · 30m solve + 5m evidence | local two-node rewiring | dummy.next points to processed prefix | cycle from pointer assignment order | O(n) | O(1) |
Hello Interview 精确目录
保留 live original titles；每行都有 canonical anchor 和 disposition。
| 项目 | 精确 canonical section | Disposition | 今天怎么用 |
|---|---|---|---|
| Ad Click Aggregator | Potential Deep Dives | deep dive | scale, no loss, dedupe, low-latency queries |
| Ad Click Aggregator | 1) How can we scale to support 10k clicks per second? | deep dive | service, stream, processor, OLAP bottlenecks |
| Ad Click Aggregator | Hot Shards | deep dive subheading | split hot AdId with suffix and merge at sink/query |
| Ad Click Aggregator | 2) How can we ensure that we don't lose any click data? | deep dive | retention, replay, raw lake archive, reconciliation |
| Ad Click Aggregator | Reconciliation | deep dive subheading | batch truth corrects low-latency speed layer |
| Ad Click Aggregator | 3) How can we prevent abuse from users clicking on ads multiple times? | deep dive | signed impression ID as idempotency key |
| Ad Click Aggregator | But how do we dedup? | deep dive subheading | pre-stream dedupe across window boundaries |
| Ad Click Aggregator | 4) How can we ensure that advertisers can query metrics at low latency? | deep dive | rollups, pre-aggregation, OLAP trade-offs |
| Ad Click Aggregator | Final Design | synthesis | speed layer + batch layer + dedupe + OLAP |
| Ad Click Aggregator | What is Expected at Each Level? | level expectation | calibrate Senior+/Staff depth |
| Ad Click Aggregator | Mid-level | level expectation | pre-aggregation basics |
| Ad Click Aggregator | Senior | level expectation | real-time processing and fault tolerance |
| Ad Click Aggregator | Staff+ | level expectation | teach trade-offs across batch/stream/storage |
DDIA 精确决策卡
只读 Ch8、Ch12、Ch13 中能改变今天设计选择的小节。
| 项目问题 | 精确 DDIA subsection | 改变的系统决定 |
|---|---|---|
| Duplicate click across minute boundary must not count twice | Ch12 · 幂等性 | signed impression ID is deduped before the stream and reconciled later |
| Flink state or consumer dies during aggregation | Ch12 · 失败后重建状态 | replay retained raw events and restore dedupe/aggregate state |
| Late events and bad deploy corrupt real-time views | Ch12 · 处理滞留事件 | admit lateness policy plus batch true-up; do not call the speed layer final |
| Batch layer must repair speed-layer mistakes | Ch13 · 应用演化后重新处理数据 | raw event lake enables reprocessing after schema/code fixes |
本日项目 artifacts
今日产出、修复与英语 recall
| 必须交付 | Ad Click deep-dive sheet：event-time window, watermark/late policy, dedupe TTL, hot shard suffix, reconciliation correction protocol。 |
|---|---|
| 修复规则 | 如果回答只说组件名，重写为 decision -> mechanism -> invariant -> failure window -> recovery -> metric。 |
| English recall | Correctness is the durable source of truth; low-latency status, aggregates, dashboards, and alerts are views with freshness and repair semantics. |
Detailed lecture notes, audio, recall scripts, PDFs, Staff Q&A, and mock packs are archived locally and are intentionally not published on this site.