Today we shipped the biggest update to Niche Sonar since launch — a set of features built around one idea: make it effortless to go from "interesting problem" to "I could build this." For the first time, the full workflow from discovery to decision is baked into the app. You don't have to export a CSV, open a spreadsheet, or manage a separate document. It all lives in one place.
What's new
Like & Save on every opportunity. Tap ♥ to signal what resonates, or ★ to save a card for later. Likes are lightweight—they're a signal to yourself and to us about what you care about. Saves are permanent: your saved cards live on a new Favorites page. Technically these are different: likes are ephemeral engagement signals we use to train ranking, saves are persistent bookmarks. In practice it means you can mark something as "that resonated" without committing to it yet.
Favorites that stick. This was important to get right. Our rankings refresh every week as new data comes in, new opportunities surface, and old ones age out. But here's the thing: if you saved an opportunity a month ago because it looked real, you want to see exactly what you saw, not the updated-today version. So every saved card carries its full snapshot—the analysis, the buyer persona, the signals—frozen in time. An opportunity you saved in May is still there in June exactly as you found it, even after the core rankings table moves on. This matters because it means you can compare notes later without chasing a moving target.
One-click AI solutions. On any saved card, hit Get Solution and you'll get a short, concrete product idea plus an honest difficulty rating (Easy / Medium / Hard) for shipping an MVP solo. It's meant to be skimmable in 30 seconds — not a business plan, just a starting point for "what would I actually build here?" The solution names the specific buyer (e.g., "freelance SEO agencies"), a one-line product idea, the GTM approach (newsletter, SMB sales, direct), and whether this is a straightforward build or a "you'll need this specialized skill" situation.
A dev blog. You're reading it. We're committing to publishing build notes and changelogs as we ship. No marketing-speak, just what we built and why.
Why both likes and saves?
Early data showed people want two different interactions. Likes let you mark something as "interesting, trend data" without friction—you skim, you see a problem that matches your experience, you tap ♥ and move on. Saves are intentional: "I'm actually considering this." Conflating them loses signal. So we kept them separate and made them visible in different places. Your likes inform your recommendation feed. Your saves are your build list.
A couple of engineering notes
Solutions are generated once, then shared. The first time anyone asks for a solution to a given problem, we generate it using Claude and cache it — keyed by a normalized version of the problem itself. Everyone who asks afterward gets that same answer instantly, with zero latency and no API cost. It keeps responses snappy and our LLM costs predictable, and it means popular problems effectively get a "community" answer that improves over time (we retrain cached solutions based on usage patterns). This is a simple pattern but it's load-bearing: without it, every AI generation is a 3-5 second wait and a $0.15 API call.
Trial users get a real taste. During your 7-day trial you get 3 free fresh AI solutions to try the feature properly; cached ones are always free to view. Paid plans are unlimited. This matters for conversion: a trial user who generates one solution probably wasn't serious; one who uses all 3 is actively evaluating. And most of our trials hit the cache anyway—the most popular opportunities get cached fast.
Knowing who's around, without sessions. We wanted lightweight "active users" stats for the public dashboard, but our auth is stateless (JWT) — there's no session table to count. So we track a throttled "last seen" timestamp on any user action (page load, card click, save), which powers both the weekly summary email and an at-a-glance "X people reviewing ideas today" metric. It's coarse but good enough, and it costs almost nothing to maintain.
What's next
More signal on each opportunity (competitive positioning, team skill requirements), and deeper solution breakdowns for those who want to go past the 30-second version. Integration with your own Notion/Linear workspace to keep build plans sync'd. As always — if there's something you want to see, hit Support and tell us.