940: Picking a Fullstack Stack, Is Next.js Too Complex? Services vs Self-Hosted + More

Episode Summary

Executive Summary: This Syntax potluck centers on practical web stack choices for Node backends, database indexing, project planning, career motivation, hosting/cost tradeoffs, REST vs GraphQL vs RPC, Vercel/Next.js skepticism, TanStack Form’s value, and a final ethics/UX hardware pick. The hosts repeatedly favor simple, standards-based tools, careful early schema/index planning, and avoiding free-tier/vendor lock-in or overcomplicated abstractions.

Main Topics: Node.js backend framework choices: For simple CRUD APIs, the hosts prefer modern, standards-based Node frameworks like Hono or H3, with Fastify and Polka as viable alternatives, because they align with web request/response conventions and are easier to move across frameworks. Database indexing strategy: Indexes should be chosen based on actual query patterns, ideally during schema design; slow-query monitoring and occasional AI-assisted review can reveal missing indexes, but not every field deserves an index due to storage and memory tradeoffs. Workflow for starting new projects: Their approach is data-first for CRUD apps: define schema, set up auth/database foundations early, create route structure, then build features incrementally front-to-back rather than scaffolding everything upfront. Career burnout and motivation: They argue that stagnation is often solved by novelty—side projects, new technical problems, hack weeks, or non-work factors like exercise—rather than by merely forcing more routine work. Choosing a SaaS stack and avoiding lock-in: For small SaaS MVPs they advise using stable, in-demand full-stack JS tools plus pragmatic external services, while being wary of generous free tiers that later become expensive or constraining. REST vs GraphQL vs RPC: GraphQL is seen as still useful for complex multi-source or multi-client environments, but too heavy for many apps; RPC-style server functions are increasingly preferred for modern full-stack web apps. TanStack Form and form complexity: TanStack Form is praised as a strong, typed, headless solution for complex React forms, though one host questions whether it solves a real need beyond what can be done with simpler form patterns and state binding.

Key Arguments: Modern Node backends should favor standards-based request/response handlers so code can move across frameworks like Hono, H3, Next.js, and Svelte with minimal rewrites. Indexes should be created from query intent, not after the fact; if you query by email, show number, or other fields often, index them early or monitor slow queries to catch misses. Not every query needs to be fast: rare admin/reporting queries may be acceptable even if slow, because index storage and memory are real costs. Starting a CRUD app should begin with data modeling and foundational concerns like auth and database shape, then proceed feature-by-feature from core to UI details. Novelty is a major driver of developer motivation; hack weeks, side projects, and new problem domains can combat burnout better than repeating the same work. Avoid building your own CMS unless absolutely necessary; use established CMSs or integrations so the product is usable by non-developers and maintainable long term. For SaaS MVPs, it is better to pick a stable stack and reliable services than to chase the cheapest/freeest combination, because vendor changes can quickly raise costs or force migrations. GraphQL is best when a single API must unify many services or clients, but for most app teams it adds too much overhead compared with REST or RPC. Vercel is not the only easy hosting option; Cloudflare, Railway, Fly.io, and even self-managed VPS/Coolify setups can provide good developer experience without the same walled-garden tradeoffs. TanStack Form may be valuable for highly complex typed form workflows, especially in React, but simple forms often do not require that level of abstraction. The episode’s broader philosophy is to prioritize simplicity, portability, and practical maintainability over fashionable or highly opinionated stacks.

Data Points: Potluck episode frequency: once a month - The hosts mention they do a potluck monthly and often have enough questions to consider doing two per month. Polka size: 90 lines of code - Polka is described as extremely lightweight, and SvelteKit uses it underneath the hood. Observed support window: 10 years - One host notes they used React for about a decade before moving on. Scope of enterprise API use case: 30 different teams - GraphQL is described as especially useful in large organizations needing a single API across many teams and services. Number of services in a no-code stack example: $700 a month - A story is referenced about someone clipping together many services and paying about $700 monthly. Cost threshold for Cloudinary: $100 a month - Cloudinary is described as generous until usage jumps to around $100/month. Sentry setup detail: slow query alerts - A missing index on the Syntax site was discovered through Sentry's slow query feature.

Pivotal Quotes: "that free tier will come and bite you in the butt once they pull the rug from you" — Host: Used while discussing SaaS infrastructure and the dangers of relying on generous free tiers. "Novelty at the end of the day is always going to be more exciting" — Host: Explaining how developers can escape burnout and stagnation through new challenges and side projects. "The new standard is web request, web response" — Host: Discussing why modern backend frameworks should align with standard web handler conventions rather than Express-era patterns.

Implications: Listeners should favor portable, standards-based stacks, design schemas and indexes intentionally, and resist over-optimizing for trendy tooling or free tiers. The industry is moving toward simpler full-stack RPC-style patterns, but long-term maintainability still depends on pragmatism and avoiding lock-in.