I built a multilingual news aggregator for European tech in 6 languages (EN, DE, FR, ES, IT, NL), filterable by audience (consumers, businesses, government) to help users discover relevant European alternatives.
A few interesting technical challenges:
PATTERN-BASED IMAGE GENERATION Generic AI images all look the same ("holographic businessman with laptop"). I solved this with a JSON database of 60+ concrete visual patterns—context-aware selection prevents repetition. Funding news → coins/contracts, security → locks/shields. Much more visual diversity. Still needs tweaking though.
GRADUAL SITEMAP GROWTH FOR NEW SITES Daily news pipeline generates lots of content. Challenge: Google indexes new sites slowly—exposing all 2K+ URLs at once resulted in 1-3% indexing rate. Solution: Dynamic rolling window in sitemap—only expose top 50 most recent articles (300 URLs), gradually increasing to 100 → 150 → 200 over 6 months as site builds authority. Result: Indexing rate improved to 10%.
AUTOMATED TRANSLATION PIPELINE RSS aggregation → AI summaries and translation with context-aware prompts → human review queue. Still tweaking quality per language pair.
STACK: Next.js 15 with Turbopack, PostgreSQL, deployed via Sliplane@Hetzner which works pretty ok.
STATUS: 1+ month post launch, 80 pages indexed, 100+ European apps featured (Mollie, Wire, Ecosia, LanguageTool, etc.)
FEEDBACK WELCOME ON: • European news sources I'm missing? • Better approaches to multilingual SEO? • Translation quality red flags? • How to optimize for LLM discovery (ChatGPT, Claude, etc.)?
Things that I liked:
* The article selection (which is probably the most important part)
Things that I did not liked:
* AI Images. While they are well made and not as cringe as others, I prefer to get stock images even if they are repeated.
* Lack of RSS (or at least I couldn't find it). This is a deal breaker for me
Nice busy!
Thanks for the feedback. Really appreciate that the article selection resonates as that's the core value prop.
RSS: You're the second to mention this already, makes it even more priority.
AI Images: Noted. The pattern was to avoid generic stock "businessman shaking hands" photos, but clearly it's not working for everyone. Options i consider now are: 1. Hide images in RSS/reader mode 2. Stock images as fallback 3. User preference toggle
Would a "reader mode" without images help?
Nice idea, but it does not work without JS. It's baffling to me that more and more sites use this (dark?) pattern of requiring JS just to display links.
EDIT: Well I was too quick, it partially works, you can click the grey "source" link on white bg for each article. Why not make the whole div a link instead ?
The sorting menus on top of the page are not working without JS though.
Thanks for testing! The whole card IS clickable actually, it has two click targets: Click anywhere on the card and it goes to the summary article page on europedigital.cloud (with translations, context, etc.) Click the "source" link and it goes directly to the original article
This way users can choose: read our summary/translations, or jump straight to the source. Tried to make it intuitive without cluttering the UI.
You're right about the sorting/filtering requiring JS though - it's client-side state. Could move that to URL params for better progressive enhancement.
Appreciate the detailed feedback, this kind of testing helps! :)
> Thanks for testing! The whole card IS clickable actually, it has two click targets: Click anywhere on the card and it goes to the summary article page on europedigital.cloud (with translations, context, etc.) Click the "source" link and it goes directly to the original article
The problem is that you're handling that with JavaScript, instead of adding a <a> with a href, and also missing href for the <img> element.
When you do this, you break the expectations of a typical web user. For example, I cannot ctrl+click or middle-mouse+click on anything of what you've done, because you personally didn't think about how people browse websites. That's not a problem, and not really your fault, lots of courses and tutorials get this wrong, teaching people to use divs then use JS for actually handling the navigation.
If instead you use a.href and img.href, you get all of these expected behavior for free, without doing anything! And as a extra bonus, people who are blind or otherwise need assistance tooling for browsing the web, will now also be able to use your website with their tooling! :)
Very good point. I'm using Next.js link components but handling the dual target behavior with JS click handlers, which does indeed break native browser behavior.
You're right that this should be: - Proper <a href> for the card wrapper - Nested <a href> for the source link - Let the browser handle ctrl+click, middle-click, etc.
Will add this to the refactor list right away. Thank you for the detailed breakdown this is exactly the kind of extra eyes and explanation I need!
Really like this. Eager for the RSS to be launched so it can be in my daily news routine.
Thank you! As of now RSS has even more priority :)
Add +1 to the priority please
Ok, RSS feeds are now live in 6 languages.
Pattern: europedigital.cloud/{locale}/news/rss
Thanks for the suggestion :)
You can select by source and audience, but not by topic. I found that a bit strange, topic would be the first thing we filter by. Why don't you dress up the summarization prompt with a task to generate tags?
Good point. You can currently search for topics (search bar), but I see what you mean, dedicated topic filters/tags would be even more intuitive than having to type.
Interesting idea about using AI to auto-tag articles with topics. What specific topics would be most useful for you? (Thinking: AI/ML, Funding, Privacy/GDPR, Regulation, Open Source, etc.)
Topic filtering could work well alongside the current audience/source filters. Will add this to the roadmap for sure, thnks! :)
very cool idea, but the AI-generated images really irk me. firstly because they're AI-generated images, secondly because they're just flat-out wrong for the example i clicked on.
Take this[0],
DATENSCUNTZ-EINSTELLUNES should be DATENSCHUTZ-EINSTELLUNGEN
ALLE AKZIETTREN should be ALLE AKZEPTIEREN
you get the point, basically almost all the text is wrong...i think AI-generated images are annoying and vapid enough, but if they don't even show actual text then what's the point? the english ones seem to be fine, can't judge the spanish ones. images feel like a lazy business-powerpoint.
anyways, cool site in general :)
[0]https://europedigital.cloud/_next/image?url=%2Fapi%2Fnews-im...
True and good catch on the German text. Fair point, AI image generation struggles with non-English text rendering. This is a known issue I've been wrestling with.
Options I'm considering: 1. Text-free abstract patterns for non-English articles 2. Language-specific image validation before publishing 3. Manual review queue for non-English images
The pattern-based approach was meant to avoid generic "holographic businessman" images, but clearly the execution needs work for multilingual content.
Thanks for the detailed feedback: this is input I need. Herzlichen dank!
> Herzlichen dank!
Gern geschehen! Mach ich doch gerne :)
maybe also add a toggle, in addition to the option you choose? option 1 would probably be the best use of your time, just don't use text for non-english generated images.
i deleted the pics in the top-row with the dev-console and i personally would prefer it that way[0] (also because i'd visit the page on an eInk screen sometimes, where no pics would be beneficial anyways, especially if they're just placeholders).
I'm getting Secure Connection Failed error in Firefox. HN hug of death?
I suggest posting this in Show HN [1] if your server can handle it.
Thanks for flagging! Checking now - site loads fine for me but might be regional/SSL issue. Can you try again? Also checking server capacity.
For others having issues, direct link: https://europedigital.cloud/en/news
Appreciate the heads up! :)
Working now for me. UK, ~19:00 UTC.
Why does this have ads and tracking?
Valid question. Let me clarify:
NO ADS: Zero advertising. No ad network scripts.
TRACKING: Only privacy-respecting essentials: . Sliplane (European hosting) server logs . No Google Analytics, no third-party trackers
What made you think there's tracking? I want to fix any privacy concerns immediately. This is a European digital sovereignty project & privacy is the whole point.
Can you share what triggered the concern? (Specific script/banner you saw?) Thanks for your help.
Clicking on "Customize" on the cookie consent banner reveals toggles for the following:
> Analytics Cookies
> Help us understand how visitors interact with our website. We use privacy-first analytics.
Tracking.
> Marketing Cookies
> Used to track visitors and show relevant advertisements.
Ads and more tracking.
> Preference Cookies
> Remember your preferences like language and theme settings.
Do these actually require separate consent, or can they be considered functional?
I would expect that a European digital sovereignty project in which privacy is the whole point would not have a cookie consent banner at all, because it would simply not use any non-functional cookies that would require it. I see the cookie banner as a sort of "mark of shame" that nefarious websites are forced to wear.
Also, I recall hearing that there were plans to make highlighting the "Accept All" button above the other options illegal, because it's a dark pattern that gets people to click the highlighted option more often.
Thank you for your persistence and pushback. Despite good intentions I fell into the boilerplate trap. In the meantime I: . removed the 'mark of shame' :) . zeroed the cookies, only localstorage for umami analytics . Simple opt-out in footer, no dark patterns, just 'learn more, opt-out' . Updated the privacy policy to reflect this Your feedback made this project better. Thank you :)
It's yet more LLM shovelware from yet another account with only self-promotion activity. Why wouldn't it?
I understand the skepticism, but let me address this:
"LLM shovelware": The articles are curated from around 30 European news sources (TechCrunch Europe, Sifted, The Verge, etc.). AI is only used for: 1. Translation (EN→NL/DE/FR/ES/IT) 2. Pattern-based image generation
The curation, source selection, and quality filters are all manual.
"Self-promotion": Fair point on the account activity. I created this account specifically to share this project with HN because the community values European tech sovereignty and privacy.
Happy to answer specific questions about the implementation. The goal is NOT traffic farming, it's building a multilingual resource for European digital policy/startups.
What would make this more valuable to you? :)
RSS or/and newsletter would be nice (but manually curated, not ai summary)
RSS feed is on the roadmap. Planning to implement: - Per-language RSS feeds (/en/news/feed.xml etc.) - Manually curated (articles are already human-selected from sources)
Newsletter is an interesting idea too. Im thinking about a weekly curated roundup per segment. What would your preference be? Thank you for your feedback, well appreciated.
I'd prefer one digest of essential highlights of the week across all segments.
I like to be interested in "other segments". But if they are in a separate subscription/email - I will never prioritize to open it.
Thank you!
Really like this approach, one unified weekly digest across all segments. You're right that separate emails would just get ignored.
Planning to make RSS first and then newsletter.
Appreciate you taking the time to share specifics: helpful for prioritization
The pointless illustrations was what stood out most for me.
Interesting perspective. Can you elaborate? Like: You'd prefer no images at all (text-only)? Or the AI-generated style doesn't work for you? Or something else?
I added them because plain text feels bare, but I'm curious what would work better. Real screenshots? Article-pulled images? Or just skip visuals entirely?
Honest feedback helps, thanks!
Not directly connected to this platform, but if I see an AI-generated image against a story I automatically assume the writer has taken as much care with their text as with their image and close the tab.
I just checked out Feedly, and it seems to use an illustration from the article when it can, a topic related generic illustration as an alternative (vague picture of an aeroplane say), or nothing.
You could use it as a model.
Thank you for your suggestion, i will certainly look into it, thanks!
No images at all please. Hacker News feels so good (IMO) in part because there is very little extra fluff. No images is part of that.
Good. I welcome it. I am quite tired of 'merica.
Haha appreciate the support! It's not about being tired of US tech it's about making sure European innovation gets, at least, equal airtime :D Coverage gap is real, just my two cents to balance things out
It's more about the american politics invading every single corner of the internet. Tech is fine, it's just that it is tainted.
european artificial intelligence ftw!!!
What's the point?, its gonna turn into another slop of "we use AI to.." type of startup updates (but with less valuation, because it's EU)
let's go!!!
Thanks! Yeah, trying to bring more attention to European tech innovation. Lots happening here that doesn't get enough coverage. Any suggestions for news sources or features you'd find useful?
