AI cost of JavaScript frameworks

#582 – November 08, 2026

the same app across 12 frameworks, measuring cost, performance, and code size

AI cost of JavaScript frameworks
16 minutes by Nolan Lawson

Web development choices are often driven by gut feelings rather than data. Nolan's benchmark tested AI coding agents building the same Pokémon app across 12 JavaScript frameworks, measuring cost, performance, and code size. MPA frameworks like Astro and Hono+htmx consistently beat SPA frameworks on speed, JavaScript size, and build cost, while Angular ranked last on nearly every measure. The results suggest framework choices deserve more scrutiny, especially now that developers are no longer writing code by hand.

Don’t sacrifice quality for speed (ElevenLabs definitely doesn’t).
sponsored by Meticulous

Agents are writing code faster than ever. They even review their own work. But who’s reviewing them? One does a passable job and ships a buggy fix, the next copies it. Bugs spread faster than anyone can catch them. Leadership at Dropbox, Notion & ElevenLabs refused to trade this quality for speed. With Meticulous, every change is exhaustively verified, with zero developer effort.

Async React & progressive server actions
6 minutes by Jacob Ebey

React now supports server-run form actions directly in the component model using the "use server" directive. New hooks like useActionState and useOptimistic handle pending states and instant UI updates cleanly. Forms still work without JavaScript as long as a server action ends up as the form's action attribute. This gives React a unified way to build experiences that work before JavaScript loads and feel great after.

React.lazy, suspense and the hidden 300 ms delay
13 minutes by Matthew Wong

All the scripts for this browser extension load within 100ms, but widgets weren't appearing until 380ms. The culprit turned out to be React's built-in 300ms delay, which batches suspended content to prevent flickering. The fix was straightforward: render a placeholder directly instead of letting React suspend, so each widget appears as soon as its code arrives. This brought the first visible widget content forward by about 240ms, while reducing bundle size made no measurable difference.

How React works
10 minutes by Sina Salahshour

React renders, state updates, and hooks all follow clear internal rules — once you understand them, confusing behavior starts to make sense. This course explains the core ideas behind React in plain terms: what actually happens during a render, why state seems stale right after setting it, and why the rules of hooks exist. No source code reading required. It pairs with a companion course that goes line by line through React's code for those who want to go deeper.

Keeping TanStack query fresh across tabs
3 minutes by Idan Shalem

When you update data in one browser tab, other tabs stay out of sync until they refetch. The fix is to broadcast a signal over the BroadcastChannel API telling other tabs that a query is stale, letting each tab fetch fresh data on its own. Idan suggests sending only the query key, not the data itself, since each tab handles its own fetching logic. The same approach works for logout, clearing private data across all open tabs at once.

And the most popular article from the last issue was:

newsletters