AI NewsGo-to-marketAnnouncement
MouseCat launches Morphy on YC, says it beats Jev on fraud
MouseCat launched Morphy on Y Combinator. The Winter 2026 team says it beats Jev and Clef by 7 to 15 percent.

Why it mattersA YC team picking a fight with the two models every fraud team buys is worth watching. Every number is still the founder's own.
Fraud teams buy a model. The model scores a card payment in a few milliseconds. It says yes or no. Two names dominate the shortlist. TypeSafe's Jev is one. Clef is the other. A YC company called MouseCat says it has beaten both.
MouseCat launched Morphy on Y Combinator's launch list on 20 November. The page has 9 votes. Founder Nicholas Aldridge wrote the pitch. He says Morphy is "almost as fast as Jev". It predicts fraud, he says, "7 to 15 percent better than Jev, Clef, or the other 4 models we tried". MouseCat is in YC's Winter 2026 batch. Its tags are Compliance and Fraud Detection.
What the pitch says
MouseCat's own site names Coinbase as a customer. Will Megson, Coinbase VP of Product, is quoted on the site. He says his team is "seeing close to 90 percent analyst-agreement rates". The site also claims one customer recovered more than 1.5 million dollars in one month. It says pairing the model with a human reviewer drops review time by 80 percent. No date and no named customer are given for the 1.5 million figure.
The YC post says Morphy "reliably flagg[s] chargebacks that common tools like Riskified and Stripe are rating low risk". MouseCat says it has started rolling the product out at paying customers. No price. No team size. No round amount is on either page.
What the launch page skips
The 7 to 15 percent gap is MouseCat's own number on its own private test set. The post states this in writing. Results are "only reported on our first proprietary dataset sourced from our opted-in customers' transactions and chargebacks", it says. "Public datasets are not complex enough" to test on. So the number is MouseCat's and the test set is MouseCat's. The company says more data will come "in the coming weeks".
The pitch still reads well, because of the shape. A decision model has two knobs. How fast it answers, and how often it is right. A model that is almost as fast and more often right is the only pair that makes a buyer switch.
For a team building software, the item is a plain reminder. A vendor's own benchmark is the start of the conversation. It is never the end. A pilot on your own last 90 days of chargebacks is the only test that counts. If Morphy clears that bar on a real ledger, the shortlist changes.
Source
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.


