the biggest lesson that i took away from our jev webinar w/ @allietheicon is that jev actually enables REAL TIME inference this is because jev is fast and cheap. you can: - do content filtering on a stream in real time - add guardrails that don't really impact ux - monitor (and steer) agent trajectory w/o prohibitive cost there's a lot more to say here but i find these examples in particular very compelling curious what others are doing in real time w/ jev! webinar here icymi:

jev as a Judge proves to be a cheaper and more precise alternative to LLM as a judge for online evals. great guide from Sean and Daniel on why Jev is great for evals, an experiment vs other LLMs, and how to try this out for your agents!

jev is so exciting! so you started using jev, now you need to know 1. what decisions is jev making? (observability) 2. is it making the right ones for my application? (evals) 3. how can i help jev make better decisions? (engine) and for obs+evals+engine, you need langsmith

instead of generating text, jev from @typesafeai generates structured output this makes it great for classification tasks like model routing, tool selection/search, and guardrails of many forms! it's also ridiculously fast and cheap compared to LLMs doing the same tasks

If you’re deep in agents, this is def worth the click @sydneyrunkle walks through Jev & @LangChain Deep Agent Harness in a way that maps to how we actually build Practical intel not demo-ware 🔖 Great info if you’re thinking about Agents

right now, a lot of agents perform classification tasks / steps if you can replace those steps with a model that's optimized for classification, you can massively reduce latency and cost i think jev is re-kindling the "graph engineering" phase of agent design

i normally would not post this late, but wanted to share what @huntlovell and I have been jamming on. @typesafeai just released Jev, a new type of classification model. it's really popular right now because it's ridiculously fast and cheap (up to 200x / 400x reductions vs comparable LLMs on classification tasks). learn all about Jev and how to build it into your harness!

overheard on my ride to work this morning, guy taking a phone call on a lyft bike “yeah RL environments”… “the value of evals…” “i mean coding agents especially…” should have thanked him for planning my next few tweets

oh how i love violin plots!! working on evaluating how a task-specific router affects agent cost and performance now just waiting for data to roll in over the next few days; blog post coming soon!

LangChain's CEO says they built their gateway a year too late, and it's still costing them. I sat down with Harrison Chase, Co-Founder & CEO of LangChain, to talk about what's changed since LangChain started as a side project in 2022, why sub-agents rarely work the way people expect, and why LangChain now runs its own hiring, content, and coding work through internal agents in Slack. 00:00 Why "I use agents for everything" is a red flag in an interview 01:21 How LangChain started as a fall 2022 side project 05:18 From LangSmith to LangGraph: rebuilding as the ground kept shifting 09:06 The gateway LangChain built a year too late 13:23 Code mode: why agents writing code beats human-style tools 19:04 Should the agent run inside the sandbox, or call it as a tool? 28:31 The agent identity problem: whose credentials does it use? 45:43 Running LangChain's own GTM, content, and coding agents in Slack 48:34 The AI slop problem, and why they built a Slack emoji for it The full pod is also available on YouTube, Apple Podcasts, or Spotify.