Ideas Radar: July 7, 2026

The most buyable gap this cycle: automated invoice collections. A builder laid out the numbers — 56% of small businesses are owed money right now, averaging ~$17,500 each, and late invoices drive up to a quarter of small-business bankruptcies. The twist is why it's unbuilt: 60% of founders won't chase overdue bills because asking their own client for money three times feels awkward, not because it takes long. A robot doesn't feel awkward — polite reminder day 3, firmer day 14, final notice day 30 — and can auto-offer payment-plan downsells that turn "can't pay $5k" into "$1,250 a month." The pricing sells itself: take a percent of recovered revenue, zero risk to the client. This needs 5-6 systems talking (QuickBooks/Xero, payment links, tone escalation, multi-channel, dispute handling), which is exactly why no owner builds it themselves.
Source: Reddit

A real developer-infra gap: fresh, trustworthy LLM pricing data. LiteLLM pulls its model-cost map from a community-maintained GitHub file, so prices lag, new models take weeks to appear or never do, and numbers are sometimes just wrong until someone opens a PR. One dev worked around it by pointing LiteLLM at a self-maintained map that refreshes daily straight from each provider — and found ~340 models missing from the official map entirely, mostly fresh releases. The clean product here is multi-source cost maps with fallback: a primary URL plus fallbacks so a missing model falls through to the next source, no matter whose map you use. As teams route across dozens of models, wrong pricing quietly wrecks cost tracking.
Source: Reddit

A universal media tracker that isn't western-only. Someone wants one app to log everything they consume — TV shows, movies, books, comics — and specifically called out that existing trackers ignore non-western media. The market has Letterboxd for film and Goodreads for books, but a single cross-format log that treats anime, manga, and international content as first-class doesn't really exist. The value is in the completeness and the underserved catalog, not another siloed tracker. A clean "one shelf for everything you watch and read" with global coverage would have an obvious, passionate audience.
Source: https://x.com/versacepond/status/2073577058727260448

An Audible library exporter. A user wants a tool that reads their Audible library and outputs an index or catalog of every book they own, instead of manually building the list by hand. It's a narrow, concrete pain — Amazon gives you no clean export — and the people who hit it are exactly the heavy buyers who own hundreds of titles. A simple "connect Audible, get a searchable/exportable catalog" utility is the kind of thing that spreads by word of mouth in reading communities.
Source: https://x.com/kennylkeys/status/2073779692176372148

A "Strava for studying" streak app. Someone asked for an app like the reading app Fable, but for studying — where you build a streak of "I studied today." The habit-streak mechanic is proven (Duolingo, Strava), and applying it to a generic "did you study" log is a clean, buildable consumer product. The specificity matters: not another bloated study suite, just a dead-simple daily-streak tracker for the act of studying. Students already gamify their habits; give them the one loop they're asking for.
Source: https://x.com/kaypaystar/status/2073628654014505323

Third-party app distribution for reMarkable tablets. A user flat-out wishes there was a way to distribute third-party software on the reMarkable e-ink tablet. There's a real, locked-in community of reMarkable owners and a developer scene making tools for it, but no official channel to find and install them. An app store or even a curated sideload-manager for reMarkable would sit on top of an engaged niche hardware base that currently has nowhere to go. Niche, but the kind of gap where whoever builds the distribution layer owns the ecosystem.
Source: https://x.com/MaximeRivest/status/2073767355721621971

An X unfollower tracker that actually works. Someone wants a tool that logs their followers so when one disappears from the list, it surfaces exactly who unfollowed. Old tools in this space broke or went paywalled after X's API changes, so the need keeps resurfacing without a reliable answer. It's a well-trodden category, but "reliable, affordable, still-working unfollow tracking for X" is a genuine open slot right now. The demand signal is steady because the pain is recurring and the incumbents keep dying.
Source: https://x.com/M_Heidemann/status/2073837540944052524

An uncensored local model for dictation cleanup. A romance writer using MacWhisper plus Ollama to transcribe dictated stories got refused mid-task — the model returned "I can't create content that involves explicit or mature themes" instead of transcribing. They don't want the model to write or alter anything, just to convert spoken punctuation commands ("open quote", "open bracket") into real punctuation and strip filler words, without censoring adult scenes. There's a real gap for a small, local, uncensored transcription-cleanup model tuned for exactly this: mechanical formatting, zero generative judgment. Fiction writers dictating spicy content are a specific, underserved slice that current safety-tuned local models actively block.
Source: Reddit