Day 3 · Wed 2026-10-07 · ChatGPT CORE
ChatGPT CORE baseline: conversations, messages, generations, and durable serving path
Week 11 trains two designs with one operating spine: admission, isolation, durable work, queue/backpressure, streaming/reconnect, derived state, evals, and recovery.
3contiguous 1-D DP slots
25-35mattempt + follow-up window
Staffinvariant, failure, recovery, metrics
Exact Daily Schedule
No drift: the same six blocks run every day.
| Time | Block | Contract |
|---|---|---|
| 08:50-09:00 | Target | Write one Staff-level invariant and one failure window before reading. |
| 09:00-10:05 | Project design | 25-35m ChatGPT CORE blank-page attempt, then entities/API/high-level path. |
| 10:05-10:45 | DDIA / source grounding | Map the design to exact DDIA anchors and only direct applied-AI sources. |
| 10:45-10:55 | Spoken close | Record the architecture in English, then say the Chinese invariant aloud. |
| 14:30-16:15 | 1-D DP contiguous block | Three NeetCode slots below; each slot is 25-35 minutes plus follow-up. |
| 20:30-21:15 | Recall / artifact | Run English Senior+/Staff recall and patch the scorecard evidence. |
Canonical Hello Interview Anchors
Direct day-page anchors only. Source titles were verified live before build.
| Canonical anchor | Disposition | Design decision |
|---|---|---|
| Understanding the Problem | CORE required reading | Constrain to text chat, streamed responses, history, and durable context. |
| Functional Requirements | CORE required reading | Message, conversation, and generation behavior must not collapse into one row. |
| Non-Functional Requirements | CORE required reading | Latency, availability, and cost are negotiated per tier. |
| The Set Up | CORE required reading | State model scope and exclude training. |
| Planning the Approach | CORE required reading | Build from entities and API into serving architecture. |
| Defining the Core Entities | CORE required reading | Conversation, Message, Generation, ModelRequest, and ContextPack are separate. |
| API or System Interface | CORE required reading | POST prompt creates a run; streaming is an observation channel. |
| High-Level Design | CORE required reading | Gateway, orchestrator, scheduler, workers, stores, and stream fanout. |
| Users should be able to send a prompt and receive an AI-generated response | CORE required reading | Generation state must survive client disconnects. |
| Users should be able to view past chats and resume a conversation with context carried across turns | CORE required reading | History is stored truth; context is selected and budgeted. |
Exact DDIA Mapping
Only Ch5, Ch9, and Ch13 sections are in scope.
| Chapter | Exact section | Interview use |
|---|---|---|
| Ch5 | 持久化执行与工作流 | A Generation has durable lifecycle independent of the HTTP stream. |
Applied AI Source Gate
First-party/original sources appear only when they change a serving, eval, security, or operating decision.
No applied-AI source is introduced today unless it changes a concrete serving, eval, security, or operating decision.
21-Slot 1-D DP Runway
Today is slot 7-9 of 21, contiguous across the week.
| Block | NeetCode problem | Mode | Invariant | 25-35m follow-up | Complexity |
|---|---|---|---|---|---|
| 14:30-15:05 | Decode Ways | Prefix count | ways[i] comes from valid one-digit and two-digit decodes ending at i. | Ask for zeros and leading-zero rejection. | O(n) / O(1) |
| 15:05-15:40 | Coin Change | Unbounded min | dp[amount] is the fewest coins; initialize unreachable with infinity. | Ask why coin order does not matter for min count. | O(amount * coins) / O(amount) |
| 15:40-16:15 | Maximum Product Subarray | Dual extremes | Track max and min ending here because negative values flip sign. | Ask for zeros and all-negative arrays. | O(n) / O(1) |
Artifacts And Recall
Use these in the evening block and scorecard.
ChatGPT Chinese deep lectureChatGPT Chinese lecture audioEnglish Senior+/Staff recall scriptEnglish recall PDF
Required spoken close: explain one invariant, one failure window, one recovery path, and one metric in English. Then restate the same design decision in Chinese without adding new components.
Detailed lecture notes, audio, recall scripts, PDFs, Staff Q&A, and mock packs are archived locally and are intentionally not published on this site.