Haeyeo’s Git: Finish It All in One Go

There are moments when a price changes in another window but doesn’t show up on your screen, even though it says “Connected.” Let’s walk through real execution logs, step by step—from how to choose between polling, long polling, SSE, and WebSocket with two virtual auction windows open, to resynchronizing after a disconnect and tracking changes across multiple servers and proxies.

Has your app ever frozen even though the server returned 200 OK? This course shows you how to define API contracts that are easy to use and safe to change by examining the requests and responses actually sent, along with the app’s runtime output. It covers everything from URLs and data shapes to errors, retries, permissions, webhooks, OpenAPI, and changes that protect existing apps, across 51 lessons in 10 sections.

I’ll show you, through an actual work session building a to-do page from scratch, the process of requesting, checking, and revising that I refined after using more than 58.3 billion tokens with Codex. You’ll learn how to ask for the results you want without knowing all the code and how to directly verify the changes.

Does changing it on the screen mean it’s been saved? From server responses and data persistence to login and permissions, AI responses and performance, and validation and deployment—this is Part 2 on what else you need to decide and verify before releasing an AI-generated interface to real users.

AI-generated SQL ran without errors—but does that mean the result is correct? Learn to identify missing criteria in a row’s meaning, values, queries, aggregations, changes, and concurrent requests, as well as which data rules to communicate to AI and what evidence to use when reviewing the results.

Even if the SQL is written correctly, is your data still protected in production? Access and permissions, recovery and safe changes, scaling and replication, MongoDB, Redis, DynamoDB, and search. This is Part 2, where you’ll learn what else needs to be defined to uphold the data standards established in Part 1 in production and across other storage systems.

Build a small RAG system with three documents, then use actual execution logs to determine where to start fixing things—documents, retrieval, context, or generation—when the answers are wrong. Quantify the impact of improvements, including hybrid retrieval and reranking, multi-stage retrieval, and permission and operational checks.

Stop copy-pasting code! 🙅‍♂️ This is a "real" conceptual lecture that dives deep into everything from how agents work to their architecture and design philosophy. Beyond just learning how to build them, I will help you develop an eye for designing entire systems. Build a solid foundation here that remains unshakable amidst the technology trends of 2026! 🧠✨

The screen created by AI looks plausible, but can you actually use it from start to finish? Learn how to find missing conditions in HTML, CSS, JavaScript, state, input, and navigation—and what screen requirements to give the AI and what criteria to use when reviewing the results.

Have you ever actually run a service you built? Deployment and observability, external integrations and queues, files and extensions, and operating AI features. This is Part 2, where you learn what else needs to be decided to keep an AI-built backend serving real users continuously.

Are you overwhelmed by AI-generated backend code and unsure where to look? Learn how to identify missing conditions in requests, APIs, data, concurrency, authentication, and caching—and what requirements to give the AI and what criteria to use when reviewing its output.

Do the results of vibe coding keep turning out differently from what you intended? Learn a workflow that uses specification-driven development (SDD) and Spec Kit to define goals, scope, business rules, and completion criteria, then shares the same specifications with Claude Code and Codex to reduce repetitive revisions.

Build a customer support agent from scratch that finds and answers from documents, selects tools to look up information, and actually creates approved tickets. Verify through real execution logs that each task is completed exactly once and through to the end, even if responses are interrupted or the same request arrives again.

“Please read the spec first and note anything that hasn’t been checked.” If you find yourself repeating the same explanation every time, it’s time to leave it in the project. This is a three-step course on connecting instructions, Skills, Hooks, and MCP to Claude Code and Codex, comfortably delegating repetitive tasks while ensuring people handle the important decisions.

When AI says “implementation complete,” do you take it at face value? This Step 2 course has Claude Code and Codex handle everything from understanding the repository to implementation, testing, review, and PRs, while teaching you to evaluate results based not on completion descriptions, but on the scope of approval, execution evidence, and change risks.

The order is confirmed, but the notification is delayed. The response is faster, but if tasks disappear after a restart, where should you look first? Using real execution records from virtual orders, we’ll walk through RabbitMQ delivery and acknowledgments, deduplication, the Outbox pattern, retries, ordering, and backlogs. We’ll also establish a way to find and complete work that remains unfinished even after a fast response.

Just because my function appears in the AI app’s tool list, does that mean it really ran? I’ll build a small issue-query server with the official TypeScript SDK and verify each step through actual execution logs—from input and result contracts to HTTP connectivity, OAuth permission checks, reducing dangerous inputs, testing, and restarting.