What it is
Radaric is a feed reader built around one job: deciding which posts deserve your attention. You connect your sources, write down in plain words what matters to you, and an AI model scores every new item against that description. Each decision comes with a confidence score and a short reason. Posts it rejects are kept, so you can check what it threw out.
It launched on Product Hunt on 23 October with the tagline "AI feed reader that learns your editorial taste", and ranked #16 for the day when we checked on 24 October 2026. The maker, Kirill Oleinichenko, edits the weekly design digest Zhurnalus; the site says he built Radaric for his own curation problem and has used it daily since November 2025. It is a one-person product: the terms name him as the operator, with no company behind it.
Who it's for
The people who get the most out of this are those whose output depends on reading more than they can: newsletter writers, curators running a daily or weekly digest, analysts tracking a niche, and creators who source ideas from Telegram channels and YouTube as much as from blogs.
If you follow 20 feeds and read everything anyway, you don't need it. If you skim 200 sources every morning to find the six worth writing about, that hour is what Radaric is after. The site claims "hours of triage turn into 20 minutes of reading". That's the company's figure with no method given, so test it during your trial.
What it actually does
Sources in one place. On the Lite plan, Radaric reads RSS, Telegram (including private channels) and YouTube, where the AI reads video transcripts rather than titles. Substack, Bluesky, Mastodon, Product Hunt, GitHub and arXiv are one-click adds. You can bring your existing subscriptions over with an OPML file exported from Feedly or Inoreader. X/Twitter accounts are only on Pro. The site also lists "Event Triggers" as a source type but doesn't explain what they are, so we've left them out of this review.
Plain-language editorial rules. Instead of keyword filters, you write a brief, the way you would brief a junior researcher. Every item gets scored against it.
Reasons on every decision. Each post shows why it passed or failed. This matters more than it sounds: when the filter gets something wrong, you can see which part of your brief it misread and fix the brief.
Learning from corrections. Approving or rejecting a post trains the filter, and you can add a reason. The site says what you save and open also counts. On Product Hunt, the maker said that "a short reason on a rejection does more than a hundred silent thumbs down", and that because he wrote his rules before the filter saw any decisions, "almost no training was needed".
Story clustering. The Story Aggregator merges posts about the same event into one story and keeps a link to every source. The site's example is one event hitting 15 of your channels, and it says this works across languages. For a curator this is the biggest time saver: one story with its list of outlets instead of fifteen near-identical headlines.
Telegram safety. The FAQ says the Telegram connection is read-only and credentials are stored encrypted, which is the claim you rely on if you connect a personal account to read private channels.
Pricing (as of 24 October 2026)
Plan | Monthly | Yearly | What you get |
|---|---|---|---|
Lite | $9 | $90 | RSS, Telegram (incl. private), YouTube transcripts, one-click sources, AI filtering with feedback learning, Story Aggregator, reasoning per decision. No X/Twitter |
Pro | $59 | $499 | Everything in Lite plus X/Twitter account parsing |
Prices are from the Radaric pricing section, which says yearly Lite is "2 months free" and yearly Pro "saves $209 vs. monthly".
There is no free tier. The site gives its reason as "the filter only gets better if you use it every day", and on Product Hunt the maker added that every post the filter judges "costs me real AI money".
Two details to check before you pay:
The 7-day trial is a Product Hunt offer. In a launch-day comment, the maker said sign-ups that come through the Product Hunt link get 7 days of Lite free, no card required, then $9 a month. This isn't on the Radaric site itself, so sign up through the Product Hunt listing if you want it.
The money-back guarantee isn't worded the same everywhere. The homepage headline says "Try it for 30 days. If it doesn't save you hours, get your money back", and the maker's launch post mentions a 30-day guarantee. But on the pricing cards the guarantee appears only under Pro, and the terms (last updated March 2026) say "Annual subscriptions include a 30-day money-back guarantee". If you plan to rely on a refund for a monthly plan, email the maker first and get it in writing.
Getting started
Use the trial window properly. Sign up through the Product Hunt link for the 7-day Lite trial, and run it during a normal working week.
Export your OPML. Feedly and Inoreader both export subscriptions. Import everything, then add the Telegram channels and YouTube channels you currently check by hand.
Write the brief before you look at a single post. This is the step the maker credits for the filter working quickly. Here is an example brief for an AI-tools newsletter for solo creators (our wording, as an illustration):
Include: new AI tools or major feature releases a solo creator could use this week; pricing changes on tools creators already pay for; primary research with a practical result; funding rounds only if the product is usable today. Exclude: opinion pieces without new information; enterprise-only products; listicles; reposts of announcements older than 7 days; crypto. Prefer: the company's own announcement over coverage of it.Correct with reasons for the first three days. Every time it gets one wrong, reject or approve it with one line: "enterprise-only, no self-serve plan" or "small feature, but it fixes the export problem our readers complain about". Then re-read your brief and tighten whatever the reasons keep pointing at.
Run your old process alongside it for one edition. Triage the old way, then compare: did Radaric surface everything you picked? Did it reject anything you would have used? Check the rejected pile once a day, which is why it's kept.
Time yourself. Note how long triage takes on day one and day six. That number tells you whether it's worth $9 a month.
If you want to go further and automate the next step, from filtered feed to drafted issue, our zero-cost newsletter autopilot covers the pipeline, and the "Original Thinker" curation engine is a DIY version of the same filtering idea if you'd rather build than buy.
Alternatives
Feedly (Feedly AI, formerly Leo). Pick it if you already live in Feedly and mainly need to cut repetition. Its documentation describes content-based deduplication that hides articles with more than 85% overlapping content, on by default. Feedly's pricing page didn't load when we checked, so we've left its prices out.
Inoreader. Pick it if you want a capable reader with a free tier and rules you control. As of 24 October 2026, Inoreader Pro is $7.50 a month billed annually or $9.99 monthly, with a free plan capped at 150 feeds. It supports Telegram and Bluesky as social feeds, offers rules, filters that remove duplicates, and AI summaries and intelligence reports. The filtering is condition-based, so you define the logic yourself and it doesn't learn your taste from corrections.
Readwise Reader. Pick it if your problem is reading and keeping what you read, not triage. As of 24 October 2026, Reader is $7.99 a month and full Readwise $9.99, billed yearly, with a 30-day free trial. It takes RSS, newsletters, YouTube, PDFs and X threads, and its Ghostreader AI summarises and explains documents. It doesn't score your feed against editorial rules. It also pairs well with a filter like Radaric, as the place you do the actual reading; our AI-first research stack shows how those layers fit together.
Limitations
No free tier, and the free trial is only available through the Product Hunt link.
X is a $50-a-month jump. If X accounts are a major part of your sourcing, the useful version of Radaric costs $59 a month, not $9.
It can't yet tell original reporting from the press release behind it. Asked directly on Product Hunt, the maker said "not yet", and that this is "still a human call". Syndicated copies are merged, but you still have to find the primary source yourself.
Refund wording is inconsistent across the homepage, pricing cards and terms (see above).
One-person product. The site says early users get direct access to the developer and adds, "This won't scale." That's a plus for support now and a risk if the maker stops. Keep your OPML export so you can leave quickly.
No stated source limits, export or API. The site mentions OPML import only. The terms also note that AI decisions aren't guaranteed to be accurate.
Verdict
Worth a 7-day trial if you run a curated newsletter; skip Pro unless X is central to your sourcing.
A written brief, reasons you can correct and duplicate coverage merged into one story are what a curator needs. Time your triage before and after the trial and you'll know whether it earns $9 a month.