About · Finance Radar
One feed from 290+ sources, ranked twice a day by four AI models, with no ads, no tracking and no engagement bait.
Finance Radar started as a personal frustration with how financial news is consumed in India. Every morning, the routine was the same — open eight browser tabs, scroll through three Telegram groups, check YouTube for new takes, skim Twitter for market chatter. By the time you'd gathered your morning brief, it was lunch.
So I built the thing I wished existed: a single, calm feed that pulls from 290+ sources across news desks, independent voices, institutional research, brokers' stock notes, Telegram channels, YouTube, and Twitter — curated not by algorithms, but by whether I'd actually trust the source with my own portfolio decisions.
There's no engagement bait. No push notifications. No "trending" section designed to spike anxiety. Just updates from sources that do real reporting, real analysis, and real thinking about Indian and global finance.
Every hour, an automated pipeline fetches RSS feeds and pages from newspapers, blogs, research houses, YouTube channels, and Telegram channels. Articles are de-duplicated, grouped by similarity, and filtered through 200+ noise patterns to strip out clickbait, market-movement filler, and rewrites.
Twice a day, four AI models from different Gemini generations — 2.5 Flash, 3.5 Flash, 3.8 Flash and 3.1 Flash-Lite — rank the feed separately. Articles picked by two or more of them rise to the top: a consensus signal that something matters, not just that someone rewrote a press release.
The homepage opens with The Big Stories: the AI groups news, Telegram reports, YouTube explainers, tweets, and research that cover the same underlying event into one story, with a one-line summary. The lead story shows its best coverage up front, the next two sit side by side, and the rest are one-line briefs under Also developing. Dots show how many of the four models picked each story. Tap one and its sources open in a side drawer (a sheet on your phone), grouped by type, so the page never jumps around.
Off to the side sits This Week: India's data releases (CPI, GDP, IIP, PMI and the rest), RBI policy and the Budget, market dates like F&O expiry and holidays, results from Nifty 200 companies, and the two global events Indian markets watch most, the US Fed's decision and US CPI. On a big screen it's a column in the margin, on a laptop a row of days you can hover over, on a phone a strip under the tabs. Dates come from NSE and the agencies' own published calendars, and one click adds them to your own calendar.
No personalisation. No tracking. The same feed for everyone, every time.
Links you click carry a small “utm_source=financeradar.news” tag, so the publisher can see Finance Radar sent you. It names the site, not you: nothing about you is recorded. Every item also has a Share link that copies the story (or opens your phone's share menu) with the same tag.
The homepage blends the AI-ranked best of everything into one page; the tabs let you go deep on any single kind of source, newest first.
Finance Radar is built entirely with Python, vanilla HTML/CSS/JS, and a healthy amount of Claude. No React. No framework. No build step. The whole site is one generated page, one stylesheet and one script, served as static files via Cloudflare Pages.
The AI ranking runs twice daily using four Gemini models (2.5 Flash, 3.5 Flash, 3.8 Flash and 3.1 Flash-Lite), explicitly prompted to favour depth over speed, explanation over opinion, and data over anecdote. The ranking models see the same feed you do — they have no inside information.
A few things you might like: bookmarks on every article, report, video, tweet and quote (copy them all as a formatted list in one click) and keyboard shortcuts (J/K to move through the news, / to search, H for Home and 1–5 for the tabs).
This is a personal project by Kashish Kapoor. It's not a startup. It doesn't have funding. It doesn't need your email. It exists because I wanted a better way to read financial news, and maybe you do too.
Know a good source I'm missing? Have feedback or suggestions?
Drop me an email →