I originally built this tiny project a few years ago but ended up abandoning it. I recently revived and finished it to solve a very specific, recurring annoyance I had with my usual workflow.

Whenever I turn on my VPN, I always want a quick visual confirmation that it’s actually routing correctly before I start browsing, usually by just checking what city/country I appear to be in.

The frustration was that whenever I googled "what is my ip" and clicked a random result, half the sites were bloated with ads, required scrolling past walls of SEO text, or bizarrely buried the actual location data (or didn't show it at all).

So I revived https://whereismyip.info/

It does two things on load:

Shows your IP and exact location front and center. Checks the ASN and automatically flags if you are connecting through a known datacenter or commercial VPN. (Also, since I live in the terminal, I added a JSON route: `curl https://whereismyip.info`).

It’s obviously free, and the source is just a lightweight Flask backend with a vanilla HTML/CSS/JS frontend (using Leaflet for the map), all routed through Caddy.

I'd love to know what you think or if your VPN ASN bypasses the detection (I know there are but the list is infinite :D)

Comments URL: https://news.ycombinator.com/item?id=47790924

Points: 1

# Comments: 0

I'm a software engineer working on C++/python/robotics by day, dabbing into web apps by night. I built The Frontier (https://the-frontier.app) because the LLM market is moving so fast it's hard to tell if you're overpaying for performance.

Pricing is easy to find, but it's hard to tell if you're missing a similarly priced or even cheaper model with better performance. So I built a visualization that maps LM Arena’s Elo scores against OpenRouter’s pricing.

The main thing it does is calculate the Pareto frontier. It highlights the optimal models at each price point, so you can easily spot when a model is technically a "bad deal" compared to its peers.

The hard part: The real headache wasn't the UI, it was the messy data. LMArena names models one way (e.g. "qwen3-coder-480b-a35b-instruct"), OpenRouter another ("qwen/qwen3-coder"), and you have to deal with a mess of variants like "thinking", "instruct", "fast", or "v1.0" vs "v1". I ended up building an automated scoring system to match these models automatically so the chart stays clean without manual mapping.

I'm pretty happy with the result, I find myself surfing the frontier (literally), going up and down the frontier to find the best model for my use case and budget.

The Tech: - React + Vite - ECharts for the visualization - A daily sync to keep the chart up-to-date with new releases

I also just added Latency and Throughput metrics because sometimes latency or throughput is just as important as intelligence.

I’d love to hear what you think, especially if you spot any weird model matches (Unfortunately they still happen) or have ideas of what to add next ! I have a few ideas, like combining latency and throughput into one, or even intelligence, latency and throughput, I'll call it Wisdom :)

URL: https://the-frontier.app/

Thanks!

Comments URL: https://news.ycombinator.com/item?id=46960202

Points: 3

# Comments: 0

As you say: software engineering doesn’t go away in the age of AI - it just moves up the ladder of abstraction ! At least in the mid term :)

I paste a json on the left, a different one on the right and it shows "No differences found. The JSON objects are identical."

Also the CSV tool should probably complain when the json is not in the intended format