ConnorLive Platform was born from frustration — and a very long postmortem document. Here's the story of why we built it.
In early 2022, Connor M. Reeves was the head of platform engineering at a Series B fintech company in San Francisco. The team was shipping fast — too fast. A botched deployment on a Thursday afternoon took down payment processing for 47 minutes. The postmortem ran to 18 pages. The root cause: a configuration drift that no existing tool had caught, combined with a manual rollback process that took 11 minutes instead of 11 seconds.
Connor called his longtime colleague Sarah T. Nakamura, who had been architecting observability infrastructure at a major cloud provider. They spent the next three weeks asking the same question to every engineering team they knew: "How long does your rollback take?" The average answer was 8 to 14 minutes. The best answer was 90 seconds. Nobody could explain why it should take more than one.
David A. Okonkwo, a product leader who had shipped developer tooling to over 400,000 engineers, joined the founding team after a single whiteboard session. The three of them agreed: the problem wasn't technical incompetence — it was tooling that was never designed for the pace and complexity of modern deployment. CI/CD pipelines had evolved, but the intelligent layer on top — anomaly detection, automated remediation, live observability stitched into the deployment loop — had been left to each team to hack together individually.
ConnorLive Platform incorporated in March 2022. The first internal deployment automation prototype was running by June. The first paying customer, a mid-sized SaaS company in Austin, signed in October 2022. By mid-2023, the platform was processing over 10,000 deployments per day. Today, more than 1,200 engineering teams across 40 countries trust ConnorLive Platform to ship software without fear.
The 3am pages are still there — that's the nature of production systems. But now they're followed by a 0.3-second rollback, a structured incident summary, and a root-cause analysis generated before the on-call engineer has finished their first cup of coffee. That's what we built. That's what we're still building.
Three engineers who decided the only way to fix the problem was to build the platform they wished had existed.
Former Head of Platform Engineering at StrideFinance. 14 years building distributed systems. Connor wrote the first ConnorLive deployment engine in a rented desk at a Tenderloin co-working space. He still reviews every major architecture PR.
Previously Staff Engineer at a major cloud provider where she led the observability platform serving 2 million developers. Sarah designed ConnorLive's AI anomaly detection system and holds two patents in distributed tracing. She runs a reading group on formal verification methods.
Product leader with 12 years shipping developer tools. Previously VP of Product at DevGrid (acquired 2019). David drives ConnorLive's product philosophy: every feature must reduce cognitive load for on-call engineers. He is an avid open-source contributor and conference speaker.
Our mission: make every deployment a non-event. These five principles guide how we build, ship, and operate.
"We believe every deployment should be a non-event. Our platform exists to eliminate the cognitive overhead, fear, and toil that drains engineering teams — so developers can focus on building, not firefighting."
— Connor M. Reeves, CEO & Co-Founder
Backed by engineering-focused investors who understand what it takes to build infrastructure for the modern software lifecycle.
We write openly about the systems, tradeoffs, and lessons behind ConnorLive Platform. No fluff — deep technical content for engineers.
A deep dive into the traffic-shifting algorithms and state machine design that make 0.3-second rollbacks possible across 50,000 daily deployments.
How we built, labeled, and continuously retrain the models that power ConnorLive's AI anomaly engine — including the hard lessons from false positive storms.
The architectural journey of making Git the single source of truth for all platform state — including the edge cases that nearly broke everything.