Jev Daily

Rivet says its support stack resolves hard tickets in 5h 1m

Nathan Flurry says Rivet built its own support tooling on Rivet plus Jev, and posted a single resolution number with nothing to compare it to.

Nathan Flurry says Rivet built its own support factory on Rivet plus Jev, and that time to resolution for hard engineering tickets dropped to 5h 1m. The post says the number will continue to improve. It does not say what the number was before.

Nathan Flurry 馃敥
@NathanFlurry
X
100% human engineers doing support, with LLM assistance
Nov 18, 2026 路 View on X

The detail worth keeping is who is doing the work. Flurry describes the setup as "100% human engineers doing support, with LLM assistance", so Jev is sitting behind people rather than answering tickets on its own. That is a narrower claim than most support automation posts, and a more credible one. Nothing in the post says which part of the pipeline Jev handles, how many tickets are in the sample, or how a hard engineering ticket is distinguished from an easy one.

The before picture

In an earlier post, Flurry said Rivet was running support on Plain and asked who had built Linear for support. He called Plain the best they had tried but not there yet, and listed what he wanted from a replacement, Slack plus Discord plus GitHub, a native mobile app, a local-first web app, stability, and a Linear integration. The new system is the answer to that post, built in-house rather than bought.

For people building on it

Treat 5h 1m as a standalone figure from a vendor measuring its own stack. There is no before number, no baseline from Plain, and no third party running the same tickets through a different pipeline, so this is not a measured improvement and it is not a benchmark. It is the kind of claim that is useful mainly as an existence proof, a small team wired Jev into its support queue and is willing to put a time on the board.

The transferable part is the shape rather than the score. Humans stay in the loop on every ticket, and the model assists, which is the configuration that keeps a wrong classification from becoming a wrong answer in front of a customer.

Nathan Flurry 馃敥
@NathanFlurry
X
It's the best we've tried, but not there yet.
Apr 4, 2026 路 View on X

Know what Jev is good at before you bet on it

One email a day. Under three minutes. What builders found out about Jev yesterday.

Get the next one by email

Jev, read daily so you do not have to. The builds, the benchmarks, the criteria that worked and the cases where it lost, from the people shipping on TypeSafe AI's System One model.