Comments URL: https://news.ycombinator.com/item?id=50120555

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48234337

Points: 7

# Comments: 1

I built SiteRay (siteray.eu) to scratch my own itch. I was getting tired of manually checking WHOIS records, SSL certs, and threat databases every time I landed on a site that felt slightly "off." Phishing and scam sites look too good these days, and the manual verification process has too much friction.

SiteRay is a web platform that aggregates multiple trust vectors to generate a 0-100 reputation score. It doesn't just check standard malware blocklists; it looks at domain age, company registry information, the existence of valid emails and phone numbers, and other signals. The web app gives you a fully transparent report so you can see exactly why a site was flagged or trusted, rather than just giving you a black-box number.

Because checking a web app still requires context-switching, I also built a companion extension for Chrome, Firefox, and Firefox for Android. It acts as a visual "traffic light" (via a top/bottom colored stripe) to give you an instant read on the domain you're currently visiting.

Privacy is obviously the biggest concern with any extension that monitors your browsing. The extension is designed to only send the root domain to our API to fetch the score—no full URL paths, no page content, and no personal data.

I'd love your technical feedback on a few specific things:

- The scoring algorithm: What edge cases or false positives/negatives are you seeing? - Data vectors: Are there other signals I should be incorporating into the trust score? - The architecture and UI: Does the transparency report give you enough actionable data?

I'll be hanging out in the comments to answer questions and take notes. Thanks for taking a look!

The core of the service is a web app where you can look up domains to get a 0-100 trust score. The backend aggregates data across multiple vectors: domain age, SSL certificates, DNS records, and various malware databases.

Because constantly copy-pasting URLs into a web tool creates friction, I also built a companion Chrome/Firefox extension that queries the API and displays the score directly in your toolbar. I’d love your feedback on the web app's analysis depth, the scoring algorithm, and the architecture!

Comments URL: https://news.ycombinator.com/item?id=47424150

Points: 2

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=47337266

Points: 2

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=46598832

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=46525248

Points: 1

# Comments: 0

I was rather skeptical about this and at first, I just waved it away as an unimportant and stupid attempt which cannot work.

But it kept following me, so I did give it a try. And I must say that the results really surprised me!

The narration was actually very good, with only a few sentences that had the wrong emotion and the such. The book sample even had some nice audio effects and music!

Maybe AI narration may really improve some of the versions of already-narrated books in the future, such as Rosemary and Rue available through Audible at the moment (https://www.amazon.com/Rosemary-and-Rue-Seanan-McGuire-audio...).

For anyone who would like to hear what I heard, here is the sample I came across: https://jumpshare.com/s/ZRibsqJ22sdhYsOCnhOD

What do you guys think about this?

Comments URL: https://news.ycombinator.com/item?id=38160204

Points: 1

# Comments: 1

The strength of it is in a single button which tells the system whether you like or dislike an article. Then let the AI model to learn why that is on its own.

If you'd like to support the project, you're welcome to do so at the Patreon page this post links to.

The old prototype also works well in many cases and is still freely available for testing. Prototype currently lives at http://44.238.200.133 (it's a demo, so no domain and no SSL there, but also no real personal data required, apart from any e-mail for login and registration purposes).

2 minutes illustration video: https://www.youtube.com/watch?v=x4l0ltXHicg

1 point by martinambrus 14 hours ago [–]

Pursuing a dream, I started re-building a prototype of a machine-learning-powered RSS readers with powerful filtering capabilities.

The strength of it is in a single button which tells the system whether you like or dislike an article. Then let the AI model to learn why that is on its own.

If you'd like to support the project, you're welcome to do so at the Patreon page this post links to.

The old prototype also works well in many cases and is still freely available for testing. Prototype currently lives at http://52.36.177.138 (it's a demo, so no domain and no SSL there, but also no real personal data required, apart from any e-mail for login and registration purposes).

2 minutes illustration video: https://www.youtube.com/watch?v=x4l0ltXHicg

The strength of it is in a single button which tells the system whether you like or dislike an article. Then let the AI model to learn why that is on its own.

If you'd like to support the project, you're welcome to do so at the Patreon page this post links to.

The old prototype also works well in many cases and is still freely available for testing. Prototype currently lives at http://52.36.177.138 (it's a demo, so no domain and no SSL there, but also no real personal data required, apart from any e-mail for login and registration purposes).

2 minutes illustration video: https://www.youtube.com/watch?v=x4l0ltXHicg