930: Text Editor Keybindings, WASM Replacing Docker, LLM apathy and hosting mini apps

Episode Summary

Executive Summary: This potluck episode covers practical developer advice and tool choices: managing VS Code/Cursor shortcuts and extension sync, evaluating titles and promotions, using LLMs responsibly for SQL and backend work, avoiding decision paralysis in a fast-moving JavaScript ecosystem, when to refactor convoluted code, the role of WASM vs Docker/serverless, micro-frontend/mini-app organization, and how AI will reshape frameworks and expectations. It ends with a label-printer sick pick and a YouTube plug.

Main Topics: Editor switching: VS Code, Cursor, and keybinding sync: The hosts discuss how they moved between VS Code, Cursor, and other editors with minimal friction, emphasizing explicit keybinding files, syncing custom shortcuts manually, and dealing with marketplace/extension differences between editors. Career growth, titles, and promotion timing: A listener asks whether to raise the issue of a missing senior title after switching jobs. The advice centers on waiting until probation ends or approaching it casually and inquisitively, without making it confrontational. LLMs, SQL, and backend/database fundamentals: They argue that developers should not rely on LLMs for database logic without understanding the underlying concepts, because security, correctness, and maintainability depend on knowing what the generated code is doing. Decision paralysis and choosing stable tools: The hosts discuss FOMO in the JavaScript ecosystem and recommend favoring mature, established technologies for production work while experimenting with newer tools only in spare time. Refactoring when code becomes convoluted: They stress that if code starts feeling overly complex, that is a signal to stop and simplify or refactor rather than layering more logic on top of a brittle design. WASM as a Docker/serverless alternative: WASM is presented as a strong fit for running purpose-built compiled tooling in JavaScript environments, though not as a universal replacement for Docker or native execution. Micro-app architecture and repository organization: For collections of mini-apps, they recommend keeping tooling light: use a monorepo or folders in one repo when possible, only adding separate package management or deployment complexity when strict versioning or isolation is needed. AI and the future of frameworks and web apps: The discussion argues that AI-generated CRUD apps are now table stakes, so the bar shifts toward better UX, performance, newer browser APIs, and frameworks that are easier for LLMs to generate correctly.

Key Arguments: Editor migration is usually smoother if custom shortcuts are explicitly stored and copied across tools; relying on hidden plugins makes cross-editor communication confusing. Extension ecosystems still differ across editors, so some manual reinstalling and project-specific enabling/disabling is normal. Job titles matter less than compensation and responsibility, but titles still affect resumes and internal leveling; ask about discrepancies casually and early if needed. LLMs can help with SQL, but developers remain accountable for security, indexing, relational design, and data integrity, so foundational knowledge is non-negotiable. The JavaScript ecosystem changes faster in discourse than in production; mature choices often reduce risk and decision fatigue. If code feels convoluted, that is usually a sign to pause and refactor rather than continue building on a bad approach. WASM is valuable for running compiled utilities in JS environments, but it is not a full replacement for containers or native/GPU-based solutions. For small side projects, keeping the structure simple and low-friction often beats sophisticated micro-frontend or monorepo tooling. AI will not eliminate the need for better software; it raises expectations and makes good performance, UX, and modern APIs more important. Frameworks and tooling may increasingly be designed to match what LLMs can generate reliably, especially around docs, JSX, and Tailwind-like patterns.

Data Points: Potluck frequency: once every month - The show opens by explaining that potluck episodes happen monthly and gather listener questions. Number of main hosts: 2 - The hosts mention that they usually answer two questions in these episodes. Question about switching editors: VS Code to Cursor - A listener asks about shortcut weirdness and settings when moving from VS Code to Cursor. Time at current job before probation ends: end of this month - A listener asks whether to bring up a senior title before probation ends. Experience level: nearly 1.5 years - A listener says they are weak on database logic after about one and a half years as a web developer. Major timeframe mentioned for JS ecosystem stability: 10 years - The hosts note that React-like choices have remained dominant for roughly a decade in production. Number of side apps mentioned: about 80 - One host describes a repo of hot tips containing around eighty apps and demos. Age of using selvage denim: 20 years - A host says they have worn selvage denim for about twenty years. Wildfire-related acreage: almost an acre - A host describes installing sprinkler systems to soak nearly an acre around a cottage to reduce fire risk. Sick pick product price: $37 - The hosts mention a larger Nimbot label printer cost in U.S. pricing.

Pivotal Quotes: "you cannot put anything into your app that you don't know what it's doing on the back end" — Scott Tolinski: Advice to a junior developer relying too much on LLMs for SQL and database logic. "as soon as something becomes table stakes, the bar is going to be raised" — Scott Tolinski: Explaining why AI-generated CRUD apps are not the end state and why software quality expectations will increase. "if things are getting too convoluted or if you're doing things in a weird way, ... stop then, peel it back" — Wes Bos: Guidance on refactoring when code starts becoming overly complex.

Implications: Listeners should treat AI as an accelerator, not a substitute for fundamentals. Mature tooling, simple architecture, and strong backend/database knowledge will matter more as AI makes basic apps cheaper to produce.