Like many reporters, I receive dozens of Google Alerts, press releases and emails every day. Finding the stories that really matter often takes far longer than writing them.
So I started building my own newsroom tool.
Using OpenAI Codex for my AI-assisted software development I’ve built an application that can:
• Read my Google Alerts
• Extract potential stories
• Group duplicate reports
• Score leads by news value and accountability
• Remember stories I’ve already seen
• Save briefings and leads for future investigations
I’ve had to learn GitHub, Flask, PostgreSQL, Supabase, Render deployments, debugging production errors, performance optimisation and database design.
It has not gone smoothly as some “improvements” ended up breaking the system.
And once I started using it I quickly found that it didn’t actually do what I needed.
This week alone I’ve realised the app needs to let me read the original source emails, trace every lead back to its alert and build a searchable research library for future investigations.
The application is still very much a work in progress (version 5.0 is the next step), but it’s now reached the point where I’m starting to use it as part of my daily routine to see if it does save time and help me to uncover better stories.
Whatever the outcome, it’s been a fun way to learn AI-assisted software development by building something to solve a problem for me.
I’m interested to hear how other journalists are using AI to improve their reporting process (and not to replace reporting).
**The video sequence has been shortened – I’m using free versions of everything (apart from Codex) so this can be a tad slow to run**