Artifact: Instagram's founders shut down their AI news app after a year

What it is

Artifact was a personalised news reader built by Instagram co-founders Kevin Systrom and Mike Krieger. It used machine learning to rank articles from selected publishers, weighing "clicks, dwell time, read time and other signals" (TechCrunch, Feb 2023).

Timeline

• Jan 2023: Artifact appears as an invite-only app that needs a U.S. phone number; the waitlist reaches around 160,000 sign-ups (TechCrunch).
• 22 Feb 2023: opens to the public, "dropping its waitlist and phone number requirements" (TechCrunch). At the time it was self-funded by the founders to the tune of "single-digit millions", with a seven-person team (TechCrunch).
• 12 Jan 2024: the team of 8 announces it will wind down; posting and comments are removed at once and news reading continues to the end of February (Artifact on Medium).
• 29 Mar 2024: Yahoo's acquisition of Artifact closes; the technology is to be folded into Yahoo News and the app no longer runs standalone (Yahoo). The price was not disclosed, and Yahoo bought the tech rather than the team (The Verge).

Why it worked (as a product, not as a business)

• Launch demand from a known team. The invite gate "drove initial demand" (TechCrunch).
• People who used it liked it. The shutdown note says "We have built something that a core group of users love", and the app was named the Google Play Store's everyday essential app of the year (Artifact).
• The tech had value to someone with distribution. Systrom said he had "on the order of 10" conversations with other companies after the shutdown note, and Yahoo started exploring a deal after reading it (The Verge).

What to learn

1. "Loved by a core group" is not a market. The founders stopped because "the market opportunity isn't big enough to warrant continued investment in this way" (Artifact). Check how many people would actually switch before you build.
2. Recommendation products need scale before they shine. Systrom: "we would get to the scale where some of the things we were promised in machine learning and AI would start working, because we had just enough scale to make them work. But it was really hard to move that needle" (The Verge).
3. Shutting down in public can create a sale. They posted the note "without shopping the company to potential suitors", then buyers called (The Verge).
4. Stop early if the answer is no. In the founders' words, "often making the tough call earlier is better for everyone involved" (Artifact).

Risks and what didn't work

• A crowded category owned by giants. TechCrunch noted it had to compete with Google News, Apple News and Meta's News Feed (TechCrunch).
• Social features cost moderation. Posts and comments were the first thing cut, because that content "requires a fair amount of moderation and oversight" (Artifact).
• Famous founders are not enough. The Verge summed it up: they "could never quite get it to scale" (The Verge).
• No user or revenue figures for the live app were published in these sources, so its real size is unknown.