Give Claude real developer news before asking for an app idea

Keep a baseline before adding the feed

Thanks to daily.dev for sponsoring this video.

Luke runs a blank-context prompt and a version supplied with current daily.dev stories. The baseline helps reveal what changed when the input improved.

1. Save the original app-idea prompt and Claude’s first list.
2. Write the product criteria that matter for your actual work.
3. Collect the personalised feed through the demonstrated API route, storing the token securely.
4. Run the comparison with the same basic request and the feed as additional context.

Require a reason for the shortlist

The selected example is Agentfile, a scanner that fetches a site without JavaScript, CSS or images and scores what remains. That is a particular diagnostic, not a complete website quality score.

1. Ask for the source story behind each idea and its connection to the criteria.
2. Select a small idea with an outcome you can test.
3. Define the checks and what the score represents before building the interface.
4. Keep the original source available so the idea can be inspected rather than treated as model intuition.

Test on more than the flattering example

The video scans Excalidraw and Luke’s own SaaS, finds a bug and reviews a shareable report. His own site receives 42 in this custom nine-check, 100-point test.

1. Run the scanner on an example with known content.
2. Compare its reported findings with the raw response and the intended checks.
3. Test a second site, fix the observed bug and rerun the same cases.
4. Keep the limitations of the custom score attached to any shared report.

The takeaway

The example tests whether better inputs produce a more useful idea. A custom accessibility-to-agents score is not proof of search ranking, traffic or product demand.