KO

Learnings

Small lessons captured before they become long-form writing

2026

    • Biome: A fast JavaScript/TypeScript formatter/linter built in Rust
    • ESLint Custom Rule
      • Write checking logic using the AST, file and type information, and more to flexibly check even complex rules
      • Report violations with context.report()
    • Biome GritQL Plugin
      • Declaratively pattern-match code structures with GritQL
      • Can modify code (rewrite) as well as diagnose it => also useful for codemode
    BiomeGritQLBasic usage of GritQL in Biome (opens in a new window)eslint.orgCustom Rules - ESLint - Pluggable JavaScript Linterhttps://eslint.org/docs/latest/extend/custom-rules (opens in a new window)
    • Jev is a model that takes natural-language input, rapidly calculates multiple typed judgments together with their probabilities, and makes them available for program decision-making.
      • unstructured state in → typed probabilistic decisions out
      • A conventional LLM generates strings, so it needs to know the preceding characters before generating the next ones. Because Jev does not generate strings, it can generate structured outputs in parallel.
      • RLCD (Reinforcement Learning from Compiler Diagnostics): reinforcement learning that checks whether a model's output matches the required type and gives it a reward
      • Ultimately, the model was created to handle judgment problems with clearly defined outputs faster and more cheaply than conventional LLMs.
      • For example, the Next.js article How we closed 1,500 GitHub issues in one month (opens in a new window) describes one agent that uses structured output to make a pass/fail judgment about whether an issue can be closed and another that looks for reasons not to close it and makes its own pass/fail judgment. The issue is closed automatically only when both agents produce the same response, and Jev seems like it could be used effectively here.
    • Perhaps Jev's real value is not “replacing LLMs,” but making it possible to use AI extensively like if statements in conventional software.
    typesafe.aiIntroducing System One Models & Jev - TypeSafe AI BlogTypeSafe AI is an AI lab building machine-native intelligence infrastructure for automation, designed to make decisions within software. Try our first System One Model, Jev, in early access. (opens in a new window)nextjs.orgHow we closed 1,500 GitHub issues in one monthHow the Next.js team used an agent to research old reports and work through the issue backlog. (opens in a new window)

    • 1. Ask questions, understand, and design

      When starting something with AI, we tend to just throw it a task and say, roughly, build this for me~ Once you have prototyped it that way, confirmed that it works, and decided to build it that way, I think it is very helpful to dig into how the actual implementation works at least once. As you keep asking questions, you start thinking, huh? Maybe this should not be done this way, or would this other approach not be better? That leads you to think of a better design.

      Asking about tradeoffs is also very useful, but AI does not readily tell you about the downsides. If you ask about problems that could arise from an implementation and the tradeoffs against other alternatives, I think you can choose a better design among them.

      In the end, I think the codebase itself is the strongest harness. That is why we need to design and implement things well from the beginning and continually question and explore the actual design and implementation methods.

    • 2. Make good use of lint

      These days, one of the things I hear most often is to use lint, and lint really is the best. If a certain rule is not being followed well, adding a custom lint rule that checks it deterministically works well. Of course, custom lint is like a test, so if it becomes too implementation-specific, there is a risk that it will become outdated. This also needs to be designed and reviewed carefully, but custom lint seems to be the cheapest and most effective way to increase the probability.

      And with custom lint, instead of merely saying that something is wrong, you can write how to fix it in the message. Or you can record lint usage, see which lint rules are repeatedly violated, and use that information in reverse to strengthen the prompt for those areas.

      • FSD also seems to be one way to keep code well maintained. It is not simply that adopting FSD is good; by enforcing the direction of dependencies (custom lint can come in here too), modules can be managed independently to create code with high cohesion and low coupling. It does not necessarily have to be FSD, but I think it is one good approach. (Reviving a broken FSD architecture with AI (opens in a new window))

    • In the end, LLMs are a probability game.

      And in order to raise that probability (to bring it close to 100%, that is, to make it more deterministic), developers' work is to introduce various engineering methods and design verification.

      1. Design carefully from the start by understanding and asking questions
      2. Gradually supplement what is lacking through a deterministic verification loop (lint being the simplest yet most effective example)

      I wonder if this could create a slightly better loop.

    토스 기술 블로그, 토스 테크AI가 팀 규칙을 지키도록 하는 방법Coding Agent를 위한 Stylepack (opens in a new window)X (formerly Twitter)shadcn (@shadcn) on Xhttps://x.com/shadcn/status/2099534231114314145?s=46 (opens in a new window)linkedin.comAI 시대의 개발자는 검증을 설계한다 (2) 구조를 검증하는 Lint'아 저 컴포넌트에다 저런 조회 로직 추가하면 안되는데.. (opens in a new window)REDBEANAI와 함께하는 무너진 FSD 살리기 (ESLint, skills, agents, hooks, CI , 자동화, claude code)안녕하세요, 프론트엔드 엔지니어 이홍빈입니다.이번 글에서는 AI를 활용하여 무너진 FSD 구조를 살린 경험에 대해 공유하려고 합니다.결론 요약AI harnessing + Guardrail 조합으로 핵심 도메인의 FSD 의존성 위반 -91%, 순환 참조 -90%를 달성했습니다.CI 자동화로 가드레일을 구성하여 지속 가능한 시스템을 구축했습니다.팀원과 함께 문제의식을 갖고, 시각화하고 공동 작업할 수 있는 기반을 마련하였습니다.지표시작 (Apr)결과 (May)감소율총 린트 에러20113-93.5%Boundary 위반14913-91.3%순환 참조313-90.3% [HTML 리포트를 활용한 과정 시각화]배경최근 프론트엔드에서 모듈을 도메인을 기반으로 설계하고, 의존성에 따라 계층화하는 FSD(Feature-sl.. (opens in a new window)GeekNewsJev - 문장 대신 판단과 확률을 반환하는 AI 모델 | GeekNewsTypeSafe AI가 분류, 라우팅, 점수화와 조건 분기처럼 소프트웨어 안에서 반복되는 판단을 자동화하는 모델 Jev를 얼리 액세스로 공개함 문장을 자유롭게 생성하는 대신 미리 정한 형식의 값과 확률을 반환하며, 여러 판단 결과를 한 번의 요청에서 병렬로 출력함 각 답에 신뢰도와 불… (opens in a new window)YouTube[한영자막] 슬롭에는 슬롭으로 맞섭니다 — Vaibhav Gupta, BoundaryTech Bridge
     (opens in a new window)    BoundaryBAMLThe programming language for agents. (opens in a new window)typesafe.aiIntroducing System One Models & Jev - TypeSafe AI BlogTypeSafe AI is an AI lab building machine-native intelligence infrastructure for automation, designed to make decisions within software. Try our first System One Model, Jev, in early access. (opens in a new window)nextjs.orgHow we closed 1,500 GitHub issues in one monthHow the Next.js team used an agent to research old reports and work through the issue backlog. (opens in a new window)NotionCustom ESLint ratcheting: how Notion modernizes code without slowing down developmentDiscover how Notion's innovative ratcheting system with custom ESLint rules gradually modernizes code, eliminates technical debt, and maintains high quality standards without sacrificing development speed or overwhelming engineers (opens in a new window)GitHubGitHub - justjake/eslint-seatbelt: Gradually tighten ESLint rules in your codebaseGradually tighten ESLint rules in your codebase. Contribute to justjake/eslint-seatbelt development by creating an account on GitHub. (opens in a new window)eslint.orgBulk Suppressions - ESLint - Pluggable JavaScript Linterhttps://eslint.org/docs/latest/use/suppressions (opens in a new window)
    • Turbopack automatically finds an optimal chunk configuration by considering download size, the number of HTTP requests, cache reuse, page navigation patterns, and other factors together.
    • Ultimately, depending on chunk size, there is a tradeoff among the number of requests, cache reuse, and file size, and it uses probabilities to find the optimal combination.
    • You can optimize the first page (firstPageLoadPriority) or specific routes (priorityRoutes), and clusters can be used to assign a higher likelihood to navigation between particular routes.
    • generateComponentChunks can be used to make fetching decisions at runtime.
    nextjs.orgHow Turbopack chunks your JavaScripthttps://nextjs.org/blog/turbopack-chunking (opens in a new window)
    • Vercel supports a Rolling Release feature as shown above.
    • React 19.3 was released
      • <ViewTransition>: Manages screen transition animations by connecting React transitions with the browser's View Transition API
        • addTransitionType() lets you specify animations for different situations
        • Using it with Suspense makes it possible to animate the transition from fallback -> content
        • Suspense can even be used for image/font loading
      • Fragment Refs: Amazing that fragments can be controlled with refs too
      • use(browser()): Suspense-based Browser-only handling
    react.devReact 19.3 – ReactThe library for web and native user interfaces (opens in a new window)