CRV

Yes, I will be at Blackhat. Yes, my calendar is a disaster. No, I have not learned any lessons from the past 10 years. Yes, I have regrets.

In Part 4 of our Building Ask DoorDash engineering series, we share how we designed the shared platform underneath Restaurant, Grocery, and Reservations — and the tradeoffs behind deciding what to centralize versus what each domain should own. A multi-agent assistant only feels like one conversation if the infrastructure underneath is shared. Ask DoorDash routes each turn to a specialized domain agent, but every agent runs on the same execution path: a Gateway that streams responses to clients, an Orchestrator that routes and pins conversations across domains, a shared MCP layer with 60+ tools that enforces permissions and business logic outside the prompt, and common services for tracing, model access, and state. That architecture changed how fast we could move: - Reservations, our third domain agent, launched in one week — roughly 10x faster than building Restaurant and Grocery - Skill-based context loading cut median instruction tokens per turn by 50%+ for Restaurant and 60% for Grocery, versus loading everything up front - Shared tracing saves roughly a month of observability work per new agent launch - A standardized model-access layer let us evaluate and deploy a new LLM within a week of release, cutting p50 turn latency by 35% with no drop in quality — and a follow-up upgrade cut it by another 40% The post covers how we decided what belonged in the platform versus what stayed domain-owned, how the Gateway and Orchestrator work together to keep cross-domain conversations coherent, and how shared state through Managed Agent Services keeps sessions, memory, and artifacts reliable across the system. https://lnkd.in/eVuadpK9

A few weeks ago, one of our reps posted publicly that working next to a teammate has been "the best apprenticeship" of her career. The teammate was Jack Buckmelter. He joined Outtake in January 2025 as our 9th team member & founding Account Executive. Zoe Bank described how, for Jack, mentoring and doing the work are the same thing. He brings her into his calls & thinks through a live deal out loud with her. She called him "the kind of guy who makes everyone around him better at sales," and that's probably why she wrote about him without anyone asking her to. Any company can put "people who look out for each other" in a values deck. Whether it happens comes down to individual choices, made every single day. Jack's made that choice since the day he walked in, and he hasn't stopped. This picture is from when Jack won our Sales MVP award. Congrats, Jack. And Zoe, thanks for putting words to something that's easy to feel around here and hard to describe.

Nuestros robots ya operan dentro de la cadena logística de MANGO 🦾 Empezamos THEKER Robotics con una idea simple: construir robots inteligentes capaces de adaptarse a entornos reales, no al revés. Hoy anunciamos nuestra colaboración con MANGO para desarrollar la próxima generación de automatización inteligente aplicada a la logística. Es tecnología en producción real, dentro de una de las cadenas logísticas más exigentes del sector retail, que abastece a más de 120 mercados. Validar esta visión junto a un referente internacional como MANGO nos acerca un paso más al objetivo que nos marcamos desde el primer día: liderar la robótica a nivel global desde Barcelona. Vamos a ganar. Jochen Grosspietsch Jordi Torra Marín Marta Mateu Vilalta

Proud CRV co-led Lotus AI's Series A and that our firm also led Outtake’s Series A, sits on its board and backed Outtake during its Series B. Now both companies are Forbes Next-Billion Startups! Congrats to KJ Dhaliwal, Alex Arjun Dhillon, Aidan Cole and the teams at both companies.

Huge congrats to the teams at Outtake and Lotus AI on being named to the 2026 Forbes Next Billon-Dollar Startups! CRV is lucky to have partnered with both of these teams since the A. It's been a privilege to watch Alex Dhillon grow the team from 8 when we invested to the powerhouse they are today. They're tackling the biggest problems in cyber from their HQ in Williamsburg. KJ Dhaliwal is building a personalized doctor in your pocket and has unbounded ambition for how healthcare should be re-imagined in the US.