What a moat is

The reason people keep paying a subscription even when the replacement can be built in a weekend.

Every page in this directory answers a single question: can an AI agent build you a personal version you would actually use? But “it can be done” and “it is worth doing” are two different things — and the second one depends on the moat: what the company has and you cannot replicate on your own. A network of people already signed up. An archive of data nobody can rebuild. Infrastructure one person cannot carry. A regulated field where getting it wrong costs a lawsuit.

We keep the English word on purpose: it is the term people actually use. Every app carries one to three moats, drawn from a closed vocabulary of 13 entries we did not invent — it ships with the dataset, so two different pages use the same word for the same thing.

The number worth looking at is this one: 66 apps out of 996 (7%) have nothing structural defending them — only execution quality. It is the most common moat in the catalogue (549 apps) and it is also the only one a coding agent erodes with every month that passes.

The 13 moats, most common first

• Execution quality

  549 apps · 55%

  Polish, reliability, sync, import fidelity: execution, not structure.

  For example: 1Password · Adobe Acrobat Pro · Adobe Lightroom · Airtable

  execution-polish

• Infrastructure scale

  444 apps · 45%

  Infrastructure one person cannot carry: global hosting, deliverability, uptime, media pipelines.

  For example: ActiveCampaign · Ahrefs · AllTrails Plus · Amplitude

  scale-infra

• Integrations

  397 apps · 40%

  How many things it connects, and the endless maintenance that keeps every connector alive.

  For example: ActiveCampaign · Airtable · Asana · Attio

  integrations

• Collaboration

  172 apps · 17%

  It only pays off with the whole team inside: shared editing, presence, permissions.

  For example: Airtable · Asana · Canny · Canva

  collaboration

• Proprietary models

  126 apps · 13%

  Trained or frontier models, plus the compute and inference underneath.

  For example: Adobe Lightroom · Bolt.new · Capture One · ChatGPT

  proprietary-models

• Proprietary data

  113 apps · 11%

  Data you cannot rebuild: indexes, crawls, real-time feeds, archives, maps.

  For example: Ahrefs · AllTrails Plus · Apollo · Clay

  proprietary-data

• Content and rights

  107 apps · 11%

  Licensed content, media rights, template and asset catalogues.

  For example: Apple Music · Audible Premium Plus · Brilliant · Canva

  content-rights

• Compliance and regulation

  102 apps · 10%

  Regulated ground: licences, payroll, tax, KYC, GDPR — real legal exposure.

  For example: Adobe Acrobat Pro · Clerky · Deel · DeleteMe

  compliance-regulatory

• Network effects

  81 apps · 8%

  It works better because everyone else is already there: graphs, communities, audience.

  For example: ActiveCampaign · Alfred Powerpack · AllTrails Plus · Apollo

  network-effects

• Brand and trust

  75 apps · 8%

  You pay because it is that vendor, and nobody ever got fired for picking them.

  For example: 1Password · Adobe Acrobat Pro · Bitwarden · Clerky

  brand-trust

• Switching costs

  40 apps · 4%

  History, configuration and habits pile up until leaving hurts.

  For example: Asana · Attio · ClickUp · Close

  switching-costs

• Hardware

  7 apps · 1%

  Physical devices, or data only the vendor’s hardware produces.

  For example: Strava · Calm · Cronometer Gold · Fitbod

  hardware

• Marketplace liquidity

  6 apps · 1%

  Two sides that serve each other, and both of them showed up.

  For example: Crowdin · Gumroad · Shopify · Fourthwall

  marketplace-liquidity

Back to every app