LG Salles

AI, sales, marketing and development

Desenvolvimento

App ideas: when a niche actually changes the product

November 18, 2026

Video by Jason Lee · Original title: “7 Tiny Apps Making at Least $40,000/month (Vibe coding)”Subscribe to the channel

Why it matters

A revenue screenshot makes an app look like an opportunity before you understand the business behind it. That is the trap I would avoid in Jason Lee’s collection of seven small products. The interesting question is what changes when a product serves a particular group: the workflow, the output, the way customers find it, or simply the wording on its website.

The revenue figures belong to the video’s account, which references Sensor Tower and, for some examples, TrustMRR. They were not independently audited for this draft. Monthly sales, recurring revenue, and profit measure different things. The selection also gives us successful examples without showing how many similar products failed to attract customers.

For developers using AI, this matters because a working interface can arrive before the underlying assumptions have been examined. Someone still has to acquire users, handle exceptions, maintain integrations, and answer support requests. My approach would be to use these examples to investigate recurring work. The additional value here is a practical way to decide whether a niche creates a meaningfully different product, followed by tests you can run before committing to a full build.

Key points

  • 00:32 Dialog makes convenience the starting point. The translation example raises a useful question: which situation becomes easier enough to justify paying for another tool? The official listing spells the name Dialog. Its existence does not establish that every free alternative has an exploitable weakness.
  • 03:14 The product needs a demonstrable benefit. Jason describes short videos and paid amplification as acquisition methods. I would look for a task whose improvement is visible in a brief demonstration. More features may not help if the benefit remains difficult to explain.
  • 03:47 PrayLock attaches an intended habit to a recurring moment. Community positioning makes the interruption more meaningful. An adaptation would need to establish whether users welcome that interruption and continue using it once the novelty wears off.
  • 07:52 PixelPainters introduces a different operating model. This is a design subscription supported by human delivery. Its official service page describes real designers. Capacity, turnaround, and revisions therefore belong in the evaluation alongside the software.
  • 10:24 Easy Expense focuses attention on one step inside a larger workflow. Capturing receipts may offer a smaller entry point than replacing an accounting system. The useful test is whether information reaches its destination with less correction and re-entry.
  • 12:02 Vid.AI prompts a hypothesis about specialized output. Serving one video format could make quality easier to define and assess. The presenter’s explanations about supplier discounts and margins remain hypotheses; they are not verified details of the company’s operations.
  • 14:48 Upward illustrates the role of audience positioning. The video presents it as Christian dating. Its current About page welcomes a broader audience seeking commitment, marriage, and family while retaining its faith-based roots. The example reflects the positioning discussed when the video was recorded.
  • 16:26 The French course-platform example adds localization to the product. Language, support, and local operating needs can affect purchasing decisions. The video does not establish that competing platforms violate applicable rules, and local hosting alone does not demonstrate compliance.
  • 18:05 The closing advice needs an evidence step. Competition, a narrow problem, and frustration with existing tools can all guide research. The missing question is which complaint creates recurring work and which is simply an individual preference.

How to apply

Filter app ideas by what changes in the workflow

Before asking an AI coding tool to build anything, complete this sentence: “When these people perform this task, they repeatedly have to correct or work around this problem.” Use an incident you can examine. “Software for contractors” describes an audience. “Assigning receipts to jobs before sending them for approval” describes work you can investigate.

Then remove the audience’s name from your proposed landing page. Does anything in the product still distinguish it? A vertical SaaS product might encode a sector’s approval rules, required fields, integrations, or delivery format. If the experience remains identical, your specialization may live mainly in the marketing. That can attract attention, but usefulness still needs evidence.

Score evidence consistently

I would compare candidate ideas using five questions. Give each zero points for an assumption, one for a specific account from a potential user, and two for something you observed. This is a personal decision aid rather than a validated success model.

| Question | Evidence to look for | |---|---| | Does the problem recur? | Recent examples across multiple work cycles | | Does the audience need a different workflow? | Decisions, fields, or outputs that generic tools handle poorly | | Can I reach these people? | A specific channel where they already discuss the task | | Can I verify the result? | A clear difference between generated output and accepted output | | Can I keep delivering it? | Known dependencies and recorded support effort |

I would hold an idea back if the workflow difference or access to prospective users remained hypothetical. Strong scores elsewhere would not answer those two questions.

Turn the audience into a product hypothesis

Here are three unvalidated directions inspired by the examples:

  • Receipts for field-service teams: test whether job assignment and supervisor approval reduce the office team’s cleanup work.
  • Videos for software demonstrations: test whether scripts tied to actual screen recordings reduce corrections about features or interface elements.
  • Course delivery for music teachers: test whether practice assignments, audio submissions, and progress tracking improve the routine beyond hosting lessons.

These require different acceptance criteria. For receipts, I would start with an export the office can actually use. For demonstration videos, I would check that each instruction matches the recorded interface. For music teaching, I would investigate whether reviewing practice becomes easier. A niche becomes useful when it changes the work the product performs.

Prototype the uncertain step first

Collect ten representative inputs with permission and appropriate handling of the data. Run the least certain transformation through the shortest workable process. Record how many outputs were accepted, how many needed correction, and how long you spent intervening.

Those observations matter more than a polished dashboard. A receipt reader can extract every amount correctly and still create work if someone must reorganize the results. A video tool can generate footage quickly while leaving hours of editing to its customer.

Measure cost per accepted result as well: processing, retries, and human correction. A subscription bundles the bill; it does not remove the work. This is also where the distinction between a software product and a service becomes practical. If every delivery requires your judgment, include that time when comparing opportunities.

Demonstrate the complete task to the intended audience

Show the input, the transformation, and the usable result. Ask prospective users where it fits into their current process and what they would still have to redo. A positive reaction offers limited evidence. Someone opening their current spreadsheet and explaining why your export would fail gives you something concrete to improve.

Finally, choose a continuation rule before expanding the build. One reasonable example is observing the same people complete two real work cycles and return without repeated reminders. If usage depends on your presence, investigate whether the obstacle is setup, trust, or insufficient value.

That is the point at which I would decide what to build next. The seven businesses provide research directions. Your own evidence has to establish which task deserves a product.

Resources mentioned

  • Dialog — Translate Speech (paid) · Download gratuito com compras e assinatura; chamado de “Dialogue” na transcrição.
  • PrayLock (paid) · Aplicativo para iPhone com compras internas; bloqueia apps até uma pausa para oração.
  • PixelPainters (paid) · Serviço de design por assinatura para igrejas, com designers humanos.
  • Easy Expense (paid) · Digitalização e organização de comprovantes; oferece recursos gratuitos e planos pagos.
  • Vid.AI (paid) · Geração de vídeos com IA; os preços e limites apresentados no vídeo são históricos.
  • Upward (paid) · Uso gratuito com planos pagos; o posicionamento atual é mais amplo que o descrito no vídeo.
  • Teachizy — provável identificação de “Teacher Z” (paid) · Identificação inferida pelo contexto francês; confirmar visualmente no vídeo antes de publicar. Oferece plano gratuito e planos pagos.
  • Lista de 20 Tiny Apps — Jason Lee · Material gratuito anunciado na descrição; página de captura do autor. O acesso não pôde ser confirmado nesta consulta.