Find the buyer before they pick
A question like "anyone using a tool for EU VAT?" is a purchase decision in progress, and it rarely carries your name.
Nephia watches your market’s words on Reddit, X, Hacker News and more, reads the intent of every mention, and files the ones showing purchase intent under one filter.
You answer while the thread is still open, with a reply already drafted in your tone.
Watch the question, not only the brand
Buyers rarely name you before they have chosen. They name the problem, the category, or the tool they want to leave. A Query on those words, "alternative to" a rival, the job your product does, catches the thread as it is posted.
One search carries several of those words at once on the sources that allow it: 10 on Solo, 25 on Growth.
Buyers, ranked first
- Purchase intent
- Someone asking what to buy, what to switch to, or whether you are worth it.
- Leads
- The Feed’s Leads filter keeps 2 intents together: purchase intent and comparison.
- Ranked by reach
- A question with an audience sits above one nobody will read.
- What the reading costs
- The AI reads mentions at 1 credit per 50 mentions, from the same credits as the checks.
The filter
intent = purchase intent
Purchase intent, as the AI read it
From the demo account. Cairn, Payhook and every handle are fictional.
Redditr/indiehackers148 score
Is Cairn worth 4 % + 40¢ on a $9/month app?
questionpurchase intent
Prices a $9 app against the fee before committing, purchase intent asked as a question.
Product HuntProduct Hunt312 upvotes
Cairn Usage, metered billing for developers
positivepurchase intent
Your own launch; nine of the comments ask about pricing before buying.
Redditr/SaaS12 score
anyone using cairn for EU VAT? how’s the invoicing side?
questionpurchase intent
Asks about EU VAT before adopting, purchase intent asked as a question.
Questions
Where do buying questions show up?
Mostly on Reddit, in comments more than in posts, then on Hacker News threads and X. A Query reads all of them at once, and the same question seen twice is one line.
What does the AI read on each mention?
Four things. A sentiment, positive, negative, a question or mixed. An intent, a complaint, a question, purchase intent, a comparison or praise. A bucket, filed against the plain-sentence instruction you gave the Query. And a rank: what deserves an answer sorts to the top by reach, and the same post seen on two Sources is one line. Counted figures (mentions, reach) come straight from the rows; inferred ones are labelled inferred. Anything you moved by hand stays where you put it, the model never overwrites a correction. It costs 1 credit per 50 items, from the same pool as your Watch ticks, and leaving it off changes nothing else.
sentiment · intent · bucket · rank
Does it post replies for me?
Only when you press Send. It drafts, in the tone you pick, neutral, friendly or formal, for every mention worth an answer. From the inbox, on a connected X account, you press Send. Nothing leaves in your name without that click. On every other source you copy the draft and post it yourself.
drafted for you · sent by your click, on X