AI Didn’t Fix Engineering. It Exposed the Systems Behind It

Software engineering walked into 2026 thinking AI was the hero. What it learned instead is that speed without discipline just breaks things faster. That tension sits underneath Jellyfish's Jan 13 LinkedIn conversation, where Adam Ferrari and Luke Stevens slowed the hype long enough to read the gauges and admit what many teams feel but rarely say out loud.

Jellyfish is not new to pressure moments. Founded in Boston in 2017 by Andrew Lau, Philip Braden, and David Gourley, three Endeca alumni who once sold a company to Oracle for $1.1B, the platform came from a blunt realization. Sales has Salesforce. Finance has accounting systems. Engineering had vibes. In an AI-saturated world, that gap stops being cute and starts being expensive.

Adam Ferrari carries long memory and scar tissue. PhD in Computer Science from the University of Virginia. Former CTO of Endeca. Nearly eight years running engineering at Salsify. A recent stint as SVP Engineering at Starburst. Now an Advisor to Jellyfish. When Adam Ferrari says boards are more pressurized about AI than anything he has seen in decades, that is not trend watching. That is lived experience.

The first signal is context engineering. Models did not suddenly get smarter. Teams got more intentional. The edge in 2026 is not waiting for better models, it is feeding existing ones cleaner context, better documentation, and clearer intent. Bad context still produces bad output, just faster and with more confidence.

The second signal is measurement growing teeth. The 2025 DORA data showed 90% AI adoption, yet delivery predictability slipped when systems stayed frozen. AI became a mirror. Strong teams got stronger. Fragile teams got noisier. Jellyfish lives in that mirror, translating commits, reviews, deployments, and dollars into something leadership can actually interrogate.

The third signal is the quiet choke point. Code review. Luke Stevens, Director of Engineering at Jellyfish, frames it clean. Faster code creation without faster review, testing, and release does not scale. It queues. Velocity upstream becomes congestion downstream. This is not a tool issue. It is a systems issue.

Today Jellyfish runs at roughly $75M ARR, serving 500+ orgs from Mastercard to DraftKings, with $114.5M raised from Accel, Insight Partners, and Tiger Global. Its newest work does not just track AI usage. It measures what survives to prod and what shows up on the balance sheet.