frisket

Hello Frisket!

Frisket, a tiny grid-based baby, emerges from its chrysalis

When I teach investigative journalists how to use AI, we're obligated to go through a laundry list of tools and techniques.

Google Pinpoint for extracting people and organizations, Gemini Notebook (née NotebookLM) for general-purpose Q&A, Claude for Sheets (RIP, my child) for extracting structured data from text blocks, MacWhisper for transcribing meetings and interviews, OpenRefine for clustering company names, and a thousand and one ever-changing tools for OCR, scraping, data analysis... The list goes on forever, not to mention that it also changes when your audience is in São Paulo or Dubrovnik or Tokyo.

So, we need a tool. A tool that:

Earlier this year I saw a few presentations about the New York Times' Cheatsheet, an internal tool used to do all sorts of fun stuff in an investigative context. I hadn't actually seen it at that point, though, my only visions were through pitch-perfect illustrations (hi Juli!) and listening to recordings from NICAR (hi Dylan!).

I'll talk more about the open source ecosystem of journalism later, but let's just say: we needed a public, open-source version of Cheatsheet. Now I knew it was something that could exist, it was time to build it.

A few rounds of automated research and one very detailed spec document later, Claude Code and I had a great hill to climb.

Hill climbing the spec

Now after four months of steering coding agents, I'm slamming that "make public" button. The features are mostly in – instead of new bells and whistles, I now spend most of my time making Codex rewrite infrastructure to be slightly less sloppy. But I suppose putting embarrassing, awful code out there is how things work in the vibe-coding era?

Before I provide the list of what Frisket can do, I want to state that there's a Guides button up at the top that walks you through eight separate workflows.

Where to find the guide

It's difficult to know how AI can aid you in the world of investigative journalism – or even normal journalism or research as a whole – and hopefully the walkthroughs will be able to provide some inspiration. At the very least they'll be more helpful than this long list of what Frisket can supposedly do! The list:

File imports

Text analysis

Cleaning

Documents

Web and research

Other things

I'm also obsessed with having options, so almost every action that does anything important has plenty of models to choose between.

The world is your oyster, and the guide is waiting! Go ahead and install Frisket on your own machine or request access to the web version.