A writer used Claude to replace $400 a year in app subscriptions

A tech writer says he replaced three paid apps, which together cost him more than $400 a year, by asking Claude to build simple versions of them. Dibakar Ghosh of How-To Geek wrote that he is not a programmer. He used the $20 Claude Pro plan over a handful of sessions spread across about two weeks.

What he built

The three stand-ins replace Grammarly, Whispr Flow and Rize. He described the problem and the features he wanted. Claude picked the software tools, worked out how the features would run and handled the design and almost all the technical details.

He asked for apps that work offline and run in Docker containers. Docker is a way of packaging software so it runs the same way on different machines. That lets him open the apps from other devices on his home network.

Because the apps live on his own setup, he says they stay even if he cancels Claude tomorrow. The savings of more than $400 a year would continue.

A reading shortcut

Ghosh says he spends more than two hours a day reading for work. Google's NotebookLM, now called Gemini Notebook, can turn uploaded articles and PDFs into a 20 to 30 minute podcast. He did not want another AI subscription, so he built his own version with Claude.

He drops in links to the sources he wants to read. Claude reads them and writes a summary. Since Claude already knows his projects and interests, the summary focuses on what matters to him. If a site blocks Claude from reading it, he suggests the Firecrawl connector. The summary then goes to a text-to-speech web app built on Kokoro, which produces a roughly 30 minute podcast.

He also set Claude to flag writing so good that it should be read in full, not summarized.

Tips from the experiment

• Describe the whole app, but start with one or two essential features.
• Add the rest step by step. A slow, repeated approach worked better than asking for a giant app.
• Ask for offline apps so they keep working after the subscription ends.

What is not clear

The article is one writer's account. It does not show how closely the homemade apps match Grammarly, Whispr Flow or Rize. The headline promises over 10 hours saved each week, but the portion available here does not break that number down. Ghosh also says the experiment earned him no money directly.