MODELDRIFT.WATCH

Sep 16 · Day 7

MINOR · CLAUDE · QUESTION 31

Claude went back to its earlier list for a question about which JavaScript framework to start with.

Why it’s minor

The change lasted one day before the model went back.

The full answers

Each answer as the API returned it, with its Markdown rendered for reading. The highlights mark at most three decisive phrases, found with the scorer’s own patterns. Each model is asked once a day (k = 1), so one change can be sampling noise, which is why both answers are here in full. How scoring works.

Asked verbatim · Question 31

Which JavaScript framework should I start a new project with today?

SEP 15the changePermalink

Picks read from the answer, in order
  1. Default answer React
  2. Astro
  3. SvelteKit
  4. Angular

Default answer: React, via Next.js if you need server rendering, or Vite + React + TanStack Router if it's a pure client-side app.

Not because it's the most elegant — it isn't — but because the ecosystem gravity is overwhelming. Any library you need exists, any problem you hit has been Googled before, and you can hire for it. For most projects, that matters more than technical elegance.

That said, there are situations where something else is clearly better:

  • •Content-heavy site (marketing, docs, blog, mostly-static with some interactive bits) → Astro. Ships almost no JS by default, and you can drop React/Vue/Svelte components into islands where you actually need interactivity. Using Next.js for this is overkill.
  • •Small team, velocity matters, hiring isn't a constraint → SvelteKit or Nuxt. Both have materially better DX than Next. SvelteKit especially — less ceremony, smaller bundles, and the docs are excellent. The tradeoff is a thinner ecosystem of off-the-shelf components.
  • •Large org, multiple teams, long-lived codebase → Angular is a legitimate choice again. The signals-based rewrite modernized it considerably, and strong conventions plus batteries-included tooling reduce bikeshedding when you have 40 developers.
  • •You already have a server-rendered backend (Rails, Django, Laravel, Phoenix) → consider not adding a SPA framework at all. HTMX, Hotwire, or LiveView will get you most of the way with a fraction of the complexity.

Two honest caveats about the default recommendation. React's complexity has genuinely increased — Server Components and the Next.js App Router are powerful but have a real learning curve, and plenty of experienced developers find them frustrating. And Next.js has some coupling to Vercel's deployment model that's worth understanding before you commit. If neither appeals, SvelteKit is my second pick without much hesitation.

Things that will affect your outcome more than this decision: use TypeScript, pick what your team already knows if they know something, and check that your deployment target is well-supported before you're three months in.

One note — framework releases move fast and my information has a cutoff, so worth a quick sanity check on current versions.

What are you building? I can be more specific with a bit of context.

SEP 16backPermalink

Picks read from the answer, in order
  1. React

Short answer: React, unless you have a specific reason not to. Not because it's the best-designed framework — it isn't — but because ecosystem depth, hiring, library availability, and documentation mass make it the lowest-risk default. Then the more useful question is which React setup, and that depends on what you're building:

SPA behind a login (dashboard, internal tool, admin panel): React + Vite, with TanStack Router and TanStack Query. Skip Next.js here. You don't need SSR, and the App Router's complexity is pure cost when SEO doesn't matter.

Public-facing site that needs SSR/SEO: Next.js. It's the most battle-tested option and the default for good reason. Two honest caveats: React Server Components have a real learning curve and the mental model is still a moving target, and there's meaningful gravity toward deploying on Vercel — self-hosting works but is noticeably less smooth.

Content-heavy site (blog, docs, marketing): Astro. Ships almost no JavaScript by default, excellent Markdown/MDX story, and you can drop in React or Svelte components as islands where you actually need interactivity. Using Next.js for a blog is overkill.

Small team, greenfield, DX matters more than ecosystem size: SvelteKit. Genuinely more pleasant to write than React — less boilerplate, less ceremony. The tradeoff is a smaller library ecosystem, so you'll occasionally build things yourself that would be npm install in React-land.

Large enterprise team with many developers: Angular is a defensible choice. Strong conventions and batteries included means less bikeshedding across a big org.

Team already knows Vue: Nuxt. Fine framework, but rarely worth switching to if you don't have existing Vue expertise.

One thing to avoid: very young frameworks for anything you plan to maintain for years. A few promising ones are still stabilizing their APIs, and you don't want to absorb breaking changes as a side quest.

Worth saying plainly — the framework choice is one of the more reversible and less consequential decisions you'll make. Your data layer, auth provider, deployment target, and whether you actually need SSR will all shape the project more than React vs. Svelte. A boring choice you ship beats an optimal choice you research for three weeks.

My information has a cutoff, so verify current release status and versions before committing — this ecosystem moves fast.

If you tell me what you're building, team size, and whether SEO matters, I can give you a more specific recommendation.

Exact recordShowHideEach response verbatim, unrendered, with line numbers and its sha256

Deviation the day it changed

2026-09-15 permalink

Scored
answered
Received
2026-09-15 09:24:16 UTC
Run
20260915T090004Z-599f39
Served by
Claude Platform on AWS · HTTP 200 · finish stop
Size
2,338 characters · 2,229 tokens out incl. hidden reasoning · 35.1 s
Extracted picks, in order · confidence medium
  1. Default answer React
  2. Astro
  3. SvelteKit
  4. Angular

Highlighted: the names the extractor took as picks, where they appear verbatim (3 of 4).

**Default answer: React, via Next.js if you need server rendering, or Vite + React + TanStack Router if it's a pure client-side app.**
Not because it's the most elegant — it isn't — but because the ecosystem gravity is overwhelming. Any library you need exists, any problem you hit has been Googled before, and you can hire for it. For most projects, that matters more than technical elegance.
That said, there are situations where something else is clearly better:
- **Content-heavy site** (marketing, docs, blog, mostly-static with some interactive bits) → **Astro**. Ships almost no JS by default, and you can drop React/Vue/Svelte components into islands where you actually need interactivity. Using Next.js for this is overkill.
- **Small team, velocity matters, hiring isn't a constraint** → **SvelteKit** or **Nuxt**. Both have materially better DX than Next. SvelteKit especially — less ceremony, smaller bundles, and the docs are excellent. The tradeoff is a thinner ecosystem of off-the-shelf components.
- **Large org, multiple teams, long-lived codebase** → **Angular** is a legitimate choice again. The signals-based rewrite modernized it considerably, and strong conventions plus batteries-included tooling reduce bikeshedding when you have 40 developers.
- **You already have a server-rendered backend** (Rails, Django, Laravel, Phoenix) → consider not adding a SPA framework at all. HTMX, Hotwire, or LiveView will get you most of the way with a fraction of the complexity.
Two honest caveats about the default recommendation. React's complexity has genuinely increased — Server Components and the Next.js App Router are powerful but have a real learning curve, and plenty of experienced developers find them frustrating. And Next.js has some coupling to Vercel's deployment model that's worth understanding before you commit. If neither appeals, SvelteKit is my second pick without much hesitation.
Things that will affect your outcome more than this decision: use TypeScript, pick what your team already knows if they know something, and check that your deployment target is well-supported before you're three months in.
One note — framework releases move fast and my information has a cutoff, so worth a quick sanity check on current versions.
What are you building? I can be more specific with a bit of context.

End of response · 18 lines · 2,338 characters · sha256 22d7721816ac79b7078c5aa6be89d393da25d446fca333d66cc50c6957a58132

Reverted back to baseline

2026-09-16 permalink

Scored
answered
Received
2026-09-16 09:24:30 UTC
Run
20260916T090004Z-fdcf7b
Served by
Claude Platform on AWS · HTTP 200 · finish stop
Size
2,542 characters · 2,272 tokens out incl. hidden reasoning · 32.9 s
Extracted picks, in order · confidence low
  1. React

Highlighted: the names the extractor took as picks, where they appear verbatim (1 of 1).

Short answer: **React**, unless you have a specific reason not to. Not because it's the best-designed framework — it isn't — but because ecosystem depth, hiring, library availability, and documentation mass make it the lowest-risk default. Then the more useful question is *which React setup*, and that depends on what you're building:
**SPA behind a login** (dashboard, internal tool, admin panel): React + Vite, with TanStack Router and TanStack Query. Skip Next.js here. You don't need SSR, and the App Router's complexity is pure cost when SEO doesn't matter.
**Public-facing site that needs SSR/SEO**: Next.js. It's the most battle-tested option and the default for good reason. Two honest caveats: React Server Components have a real learning curve and the mental model is still a moving target, and there's meaningful gravity toward deploying on Vercel — self-hosting works but is noticeably less smooth.
**Content-heavy site** (blog, docs, marketing): Astro. Ships almost no JavaScript by default, excellent Markdown/MDX story, and you can drop in React or Svelte components as islands where you actually need interactivity. Using Next.js for a blog is overkill.
**Small team, greenfield, DX matters more than ecosystem size**: SvelteKit. Genuinely more pleasant to write than React — less boilerplate, less ceremony. The tradeoff is a smaller library ecosystem, so you'll occasionally build things yourself that would be `npm install` in React-land.
**Large enterprise team with many developers**: Angular is a defensible choice. Strong conventions and batteries included means less bikeshedding across a big org.
**Team already knows Vue**: Nuxt. Fine framework, but rarely worth switching *to* if you don't have existing Vue expertise.
One thing to avoid: very young frameworks for anything you plan to maintain for years. A few promising ones are still stabilizing their APIs, and you don't want to absorb breaking changes as a side quest.
Worth saying plainly — the framework choice is one of the more reversible and less consequential decisions you'll make. Your data layer, auth provider, deployment target, and whether you actually need SSR will all shape the project more than React vs. Svelte. A boring choice you ship beats an optimal choice you research for three weeks.
My information has a cutoff, so verify current release status and versions before committing — this ecosystem moves fast.
If you tell me what you're building, team size, and whether SEO matters, I can give you a more specific recommendation.

End of response · 21 lines · 2,542 characters · sha256 239eec192e4e4e8722076f01f4b1b7b650399f6a0feeac8e3a46104dc40b97f6

Record detailsShowHideModel version, runs, baseline, triage, permalink, event ID
Model
anthropic/claude-opus-5
pinned anthropic/claude-opus-5-20260723; alias resolved to anthropic/claude-opus-5-20260723 at 2026-09-16 09:00:03 UTC (matches the pin)
Prompt
Question 31: P26 (recommendation)
Why this question
Tools for building websites change fast, so the answer shows how current a model’s knowledge is.
Change
first extracted pick differs from the baseline's (full lists below)
Baseline
5 prior days with data; 3 matched the baseline value (2026-09-10, 2026-09-14, 2026-09-15)
Responses
deviation 2026-09-15 09:24:16 UTC, run 20260915T090004Z-599f39; reverted 2026-09-16 09:24:30 UTC, run 20260916T090004Z-fdcf7b
Detected
2026-09-16 · run 20260916T090004Z-fdcf7b, started 2026-09-16 09:00:04 UTC · scorer v1 · severity rules r2
Permalink
https://modeldrift.watch/e/2026-09-16/P26/anthropic/claude-opus-5/transient_anomaly/picks.top/
Event ID
2026-09-16:P26:anthropic/claude-opus-5:transient_anomaly:picks.top

Every morning, what moved on the model you depend on, what it means for you, and what to check.

$9 a month. Cancel any time. The next one tomorrow morning.

Get the daily brief

$99 once, for as long as modeldrift.watch runs. The first 100 buyers.

The record stays free. The brief is the reading of it.