iiinova — where your digital self keeps evolving

Jan @amann.work · 09/09/2026

Yeah, absolutely! I've heard a couple of times that people (still) struggle with this, I might look into it … I wonder if it should work similar to automated code reviews on GitHub, where the author of a PR gets notified, but can choose to fix or not fix a reported finding.

010

Jan @amann.work · 09/09/2026

Were hardcoded strings a problem you had in a codebase?

100

Jan @amann.work · 09/09/2026

It depends. There is react/jsx-no-literals, which covers this well. But many strings also live in props, where it's trickier to detect. And then you might have toast.show('Hello!'), which is also a hardcoded string. So it depends on where you draw the line, this might need AI analysis.

100

Jan @amann.work · 09/09/2026

Would you like eloqnt/cli to catch hardcoded strings in your components? Does your agent get this wrong sometimes?

100

Jan @amann.work · 01/09/2026

Seems like eloqnt/cli is helpful to you! Anything it could do that it doesn't today?

111

Jan @amann.work · 27/08/2026

next-intl 4.14 is out with 𝚞𝚜𝚎𝙴𝚡𝚝𝚛𝚊𝚌𝚝𝚎𝚍 improvements! → Switch 𝚖𝚜𝚐𝚒𝚍 of .𝚙𝚘 codec to source string → Support 𝙿𝚛𝚘𝚖𝚒𝚜𝚎.𝚊𝚕𝚕 with 𝚐𝚎𝚝𝙴𝚡𝚝𝚛𝚊𝚌𝚝𝚎𝚍 → Update 𝚜𝚠𝚌_𝚌𝚘𝚛𝚎 to 77

120

Jan @amann.work · 17/08/2026

With 𝚗𝚎𝚡𝚝/𝚛𝚘𝚘𝚝-𝚙𝚊𝚛𝚊𝚖𝚜, you can cache more of your app. If you read a locale from 𝚌𝚘𝚘𝚔𝚒𝚎𝚜() or 𝚑𝚎𝚊𝚍𝚎𝚛𝚜(), this is request data. A locale from [𝚕𝚘𝚌𝚊𝚕𝚎] is a cache key. Your routing strategy becomes a caching decision.

120

Jan @amann.work · 12/08/2026

Next.js 16.3 includes better ISR: Pages can now show an instant loading shell on first visit, and then upgrade to a fully prerendered page in the background (cached). This comes in handy for apps with a large number of locales!

183

Jan @amann.work · 06/08/2026

𝚎𝚕𝚘𝚚𝚗𝚝 𝚕𝚒𝚗𝚝 now links to rule details, so agents can get more guidance when fixing issues.

000

Jan @amann.work · 04/08/2026

Read all about it here → next-intl.dev/blog/nextjs-...

020

Jan @amann.work · 04/08/2026

Next.js 16.3 is out with next/root-params 🥹🥹🥹 I've literally waited years for this! Use it today with next-intl 🔗↓

140

Jan @amann.work · 03/08/2026

studio.eloqnt.dev/docs/lint-ru...

000

Jan @amann.work · 03/08/2026

With the number of lint rules for eloqnt/cli growing continuously, I've now added dedicated pages for each one. Is this helpful? Anything to add/change?

100

Jan @amann.work · 27/07/2026

eloqnt/studio now has built-in Crowdin synchronization

100

Jan @amann.work · 27/07/2026

MDX components in eloqnt/studio now render twice: → React for humans → Markdown for agents

110

Jan @amann.work · 27/07/2026

TIL in monorepos you can split 𝙲𝙻𝙰𝚄𝙳𝙴.𝚖𝚍 files by package, much tidier 💡 𝚎𝚕𝚘𝚚𝚗𝚝/ ─ 𝙲𝙻𝙰𝚄𝙳𝙴.𝚖𝚍 ─ 𝚊𝚙𝚙𝚜/ ─ 𝚜𝚝𝚞𝚍𝚒𝚘-𝚠𝚎𝚋/𝙲𝙻𝙰𝚄𝙳𝙴.𝚖𝚍 ─ 𝚙𝚊𝚌𝚔𝚊𝚐𝚎𝚜/ ─ 𝚌𝚕𝚒/𝙲𝙻𝙰𝚄𝙳𝙴.𝚖𝚍 ─ 𝚜𝚍𝚔/𝙲𝙻𝙰𝚄𝙳𝙴.𝚖𝚍

110

Jan @amann.work · 24/07/2026

Shipped a much-needed docs revamp for eloqnt/studio. → Zero-to-lint quickstart (with a prompt) → Copyable .𝚖𝚍 links → Support for 𝙰𝚌𝚌𝚎𝚙𝚝: 𝚝𝚎𝚡𝚝/𝚖𝚊𝚛𝚔𝚍𝚘𝚠𝚗 → Split into multiple pages → Public changelogs → It looks just a lot nicer :D

100

Jan @amann.work · 17/07/2026

When should you translate your app? After every change? Before a commit, after a commit? What about the moment you switch the PR from "draft" to "ready for review"? bsky.app/profile/aman...

110

Jan @amann.work · 16/07/2026

What makes eloqnt/cli somewhat special, is that it understands not only your i18n strings, but also your source code. Helpful for quality checks, but also for updates. bsky.app/profile/aman...

000

Jan @amann.work · 15/07/2026

Have you tried linting your i18n strings yet? > npm install @eloqnt/cli --save-dev

010

Jan @amann.work · 14/07/2026

Start linting your strings today. 𝚗𝚙𝚖 𝚒 @𝚎𝚕𝚘𝚚𝚗𝚝/𝚌𝚕𝚒 --𝚜𝚊𝚟𝚎-𝚍𝚎𝚟 Add .𝚎𝚕𝚘𝚚𝚗𝚝/𝚌𝚘𝚗𝚏𝚒𝚐.𝚝𝚜 and run: 𝚗𝚙𝚡 𝚎𝚕𝚘𝚚𝚗𝚝 𝚕𝚒𝚗𝚝 Docs → studio.eloqnt.dev/docs

100

Jan @amann.work · 14/07/2026

Localization outgrowing your repo? eloqnt/studio is modular—keep it for linting and source review, and swap out the translation layer for your favorite translation management system. Crowdin support is built in, so upgrading is a config file away—unlocking reviews, 3rd party integrations, and more.

100

Jan @amann.work · 14/07/2026

The backbone of eloqnt/studio is its AI engine. You can either use a hosted version, or bring your own model—anything supported by the AI SDK. This also includes 3rd-party plugins for using your Claude Code or Codex sub (subject to their latest terms).

100

Jan @amann.work · 14/07/2026

And it all lives in your repo. Config, styleguides, messages—versioned and branched with your code. Nothing to sync, nothing drifting apart on a feature branch.

100

Jan @amann.work · 14/07/2026

What makes your product or service unique? Terms to keep, tone to hold, a glossary—all placed in simple markdown styleguides. Easy to set up and maintain, and the format your agent already speaks. It's AGENTS.md, but for translations.

100

Jan @amann.work · 14/07/2026

After maintaining next-intl for years, I've seen every way translations break. eloqnt/studio matches strings to their references in source code. This not only enables additional quality guardrails—it pulls in all the relevant context your agent has always silently guessed at.

100

Jan @amann.work · 14/07/2026

So eloqnt/studio is different. It's not a model, but a pipeline. Every string moves through well-defined steps: ① Static analysis ② Source text review ③ Context enrichment ④ Translate

100

Jan @amann.work · 14/07/2026

AI translation is "good enough"—right up until it isn't. I kept hearing: "We have this duct-taped Claude Code skill that does it for us." "But honestly, we have no idea if the output is any good." bsky.app/profile/aman...

bsky.app

Jan (@amann.work)

AI will translate your app in seconds. It'll also confidently mistranslate, break a plural, flatten your tone—and you'll ship it, because you can't review what you can't read. [contains quote post or…

100

Jan @amann.work · 14/07/2026

Today, I'm launching eloqnt/studio. An AI-based toolchain that helps you ship high-quality translations. All from the command line. 𝚎𝚕𝚘𝚚𝚗𝚝 𝚕𝚒𝚗𝚝 𝚎𝚕𝚘𝚚𝚗𝚝 𝚛𝚎𝚟𝚒𝚎𝚠 𝚎𝚕𝚘𝚚𝚗𝚝 𝚝𝚛𝚊𝚗𝚜𝚕𝚊𝚝𝚎 With our hosted model, or bring your own. 🔗↓

110

Jan @amann.work · 09/07/2026

What if your linter caught translation bugs the way it catches code bugs? Missing translations. Unused keys. Broken ICU args. The list goes on.

050

Jan @amann.work · 02/07/2026

AI will translate your app in seconds. It'll also confidently mistranslate, break a plural, flatten your tone—and you'll ship it, because you can't review what you can't read.

010

Jan @amann.work · 01/07/2026

Seems to be what many developers are doing, from as far as I can tell! Maaaybe I have something in the works in this regard 👀

010

Jan @amann.work · 30/06/2026

How do you keep translations in sync when you're coding with agents?

220

Jan @amann.work · 19/05/2026

A new Payload CMS template is available for members of 🌐 learn.next-intl.dev — let me know if you find it helpful!

010

Jan @amann.work · 13/05/2026

🌐 next-intl 4.12 is out with useExtracted improvements! → Improved monorepo support → Stacked .po descriptions → Path-only references by default → Deferred one-off extraction → Another message ordering fix See the release notes for details: github.com/amannn/next-...

github.com

Release v4.12.0 · amannn/next-intl

4.12.0 (2026-05-13) Features Improvements for useExtracted (#2316) (bff2f96) – by @amannn ⚠️ A small config change is required for useExtracted users to upgrade, please see #2316 for details.

020