TypeSafe AI

Jev is a phenomenal reranker. 🎤 https://lnkd.in/gs8uJjpx

Excited for n8n launching TypeSafe nodes so that Jev can power decisions in your workflows! https://lnkd.in/gmiVq_ZY

TypeSafe takes Portola. 🎶 Work at TypeSafe doesn't really stop, but this weekend, part of the team stepped away to catch Portola Music Festival together. A rare pause, and a well-earned one. #portola #teambonding #typesafe #startup #workculture

We're hiring a Founding Recruiter.⚒️⚡️ At TypeSafe, we hire strong horsepower people. This is a ground-floor seat. High ownership, high trust, and a direct hand in shaping who this company becomes. If the Jevolution resonates with you, and you know how to find people who get it too, we want to talk. #sf #typesafe #tech #foundingrecruiter #jev

What color is our favorite? Wrong answers only 🩷

The closest thing to a deterministic LLM call just landed in n8n. And it's free via n8n Gateway credits until Oct 10. TypeSafe AI's Jev is the model everyone's been talking about this month. It doesn't write text. It doesn’t hallucinate. You give it an input, a question, and your desired outcomes and it returns the probability for each one. Low latency and low cost mean you can use the TypeSafe AI node anywhere a workflow needs a judgment call: → Does this deal need an exec review? → Is this ticket critical, high or low priority? → Complaint, question or spam? 🎁 On n8n Cloud (Starter, Pro, and new trials), Jev runs on Gateway credits and it's free until October 10. If you’re self-hosted or on the Community Edition, you can BYOK once the node is installed. Search "Jev" in the nodes panel (refresh your instance if you don't see it yet) to try it in your own workflows. 🎥 Want to see it in action? Check out Liam’s video below. https://go.n8n.io/FsedIr

Some roles you can explain in a sentence. This isn't one of them. 😈 We're hiring for our Data Demons (AKA Model Capabilities) team. Watch the video. Apply if you dare. #typesafe #modelcapabilities #data #jev

Our manifesto: https://lnkd.in/gQKFGUDf We hire strong horsepower people no matter the role. If the jevolution or jevons paradox resonate, this may be for you. We know we blew up - our entire team has been working extremely hard to serve Jev. Definitely want to sleep well some day. 🫠 If we have worked together, you need to join yesterday. DM me else please apply at TypeSafe.ai/careers

Up to 59× cheaper. Up to 4× faster. Quality held or improved across tasks. That's what one of the world's largest HR and payroll platforms found when it put Jev head-to-head with frontier LLMs on its own production problems. Deel got early access to Jev on Sunday. Within days, their teams had run it against the frontier LLMs they use in production today. Same tasks, same data, same eval sets. Every task had the same shape: pick the right answer from a known set of options. Across analytics, support triage, ticket classification, and expense categorization: •Cost: 20 to 59× cheaper per decision •Speed: up to 4× faster on live production traffic, in shadow mode •Quality: parity or better on 6 of 8 checks •Matching repeat questions to approved answers: 70% for the frontier LLM, 97% for Jev. •Sorting expenses into about 55 categories: 50% for the frontier LLM, 86% for Jev, measured against human reviewers. When no valid answer existed, Jev said so instead of inventing one. That is what let a team automate the confident cases and route the rest to a person. Frontier LLMs still lead on open-ended questions. Jev is built for the high-volume decisions underneath them. What parts of your business could you improve with Jev?

Model Capabilities Team, or as we call them, 😈 Data Demons. Hiring now!