The End of Manual Debugging
TL;DR
Sasabi’s core bet is “logs are all you need” — Sherwood argues that AI makes unstructured logs newly powerful, so teams no longer need to painstakingly instrument logs, metrics, and traces like the old “three pillars of observability” playbook from Datadog-era tooling.
The real pain point isn’t writing code anymore — it’s maintaining production systems — after years at Crunchbase, Brex, and 11X, Sherwood says AI coding tools transformed software creation, but debugging outages still meant hours of dashboards, log searches, and flame graphs.
Brex taught him why observability exists in the first place — as one of the earliest infrastructure engineers around employee ~70, he helped support 50 microservices in Kubernetes and saw firsthand that no amount of testing prepares you for production: “everyone has a plan until they get punched in the face.”
Opkit worked as startup training, not founder-product fit — his first YC company (Summer 2021) built voice AI and revenue-cycle tooling for healthcare, but Sherwood now describes it as an “MBA case study” idea chosen from market logic rather than personal experience, passion, or strengths.
Sasabi is explicitly designed as the opposite of Opkit — Sherwood says he wanted his second company to align with who he is: observability, dev tools, AI, early-stage startups, and even the name Sasabi, pulled from Gundam, as a symbolic act of founder-product identity.
He’s doing YC a second time to compress time-to-market and win distribution — despite already being in the network, Sherwood sees YC as a forcing function for speed, culture, and go-to-market, especially because every YC startup is a software company that already needs observability.
The Breakdown
A Datadog for 2026, Not 2012
Sherwood opens by pitching Sasabi as an AI-native observability platform for fast-moving engineering teams — “like a Datadog or a Sentry,” but built for 2026. The whole idea comes from a familiar pain: every outage meant hours lost spelunking through dashboards, log searches, and flame graphs, while AI had already started changing everything else about software creation.
“Logs Are All You Need”
His hottest take is also Sasabi’s manifesto: logs alone should be enough. Sherwood argues the old observability model — logs, metrics, and traces — created huge instrumentation overhead, while logs remain the most natural thing for developers because everyone understands a print statement and a log stream. What changed is AI: logs used to be the least machine-readable signal, but now agents can read unstructured log lines and tell you why production is down, what errors mean, which customers are affected, and even which commit likely caused the issue.
From Junior Front-End Engineer to Brex Observability Builder
At Crunchbase, Sherwood started as the most junior front-end engineer, got stuck with the unglamorous CI/CD problem, and unexpectedly fell in love with infrastructure. At Brex, joining around employee 70 as effectively the third infrastructure engineer, he helped build the staging and production environments, microservice framework, CI/CD, and eventually observability as the company grew into 50 Kubernetes microservices owned by different teams.
The Mike Tyson Theory of Production
At Brex, observability stopped being tooling and became philosophy. Sherwood says you can stack up integration tests, unit tests, QA, release processes, and static analysis, but production still surprises you — invoking Mike Tyson’s line that everyone has a plan until they get punched in the face. His team set up auto-instrumentation, Datadog, dashboards, monitors, and SLO/SLI adoption so every service could at least be understood once reality hit.
Opkit: A Smart Idea That Wasn’t “His”
Sherwood then tells the story of his first company, Opkit, from YC Summer 2021: a healthcare voice-AI and revenue-cycle business handling insurance eligibility, prior auth, and claims workflows. He chose healthcare partly because his dad was a doctor and partly because it looked like a giant vertical market, but in hindsight he says the company was picked like an “MBA case study” — opportunity-first, not identity-first.
Learning the Hard Way, Then Pivoting Into AI Voice
Opkit evolved from healthcare workflow software into one of the earliest teams commercializing LLM-based voice agents in healthcare, even running a call center in the Philippines to support insurance operations. The engineering challenge was exciting, but fundraising on the new direction was lukewarm, and with about six months of runway left they decided not to force it; instead they explored acquisitions and ultimately joined 11X, an AI sales company, through a connection to its CTO from Brex.
The 11X Moment That Made Sasabi Obvious
At 11X, Sherwood rebuilt the AI SDR product in roughly three to four months, then found himself in the same old maintenance loop: setting up Datadog again, handling on-call issues, and manually debugging problems for a futuristic AI product with decidedly non-futuristic tooling. That contrast made the opportunity click — if AI can transform code generation, it should also transform the much larger job of maintaining software already in production.
Building the “Tombstone Company” and Coming Back to YC
Sherwood says Sasabi is the first company fully aligned with who he is: AI, infrastructure, observability, developer tools, startups, even Gundam. He wants it to become the default observability tool for the AI era and eventually power “self-healing software,” while building what he calls the “Linear of observability” — beautiful, fast, and engineer-loved. Coming back to YC is less about badge value and more about acceleration, shipping cadence, and distribution: every YC company is a software company, every software company needs observability, and he wants Sasabi to be the one they use.