The pattern
The same play, four times.
Every era of IT arrived as a gift and left as a toll. Here's the pattern we keep living through — and why we're building Bifrost before it repeats. These aren't one neat evolution; they're four separate waves, each with its own savior and its own bill. The only constant is the capture.
Break-Fix
early 2000s
The lift
Remote access (LogMeIn, 2003) cut out the drive for every fix; Small Business Server standardized how SMBs ran.
The toll
LogMeIn axed its free tier overnight in 2014 — seven days to pay up. SBS was discontinued in 2012; the all-in-one box quietly disappeared.
Managed Services
2004 →
The lift
RMM and PSA — LabTech (2004), ConnectWise — turned reactive fixes into a real, recurring business.
The toll
ConnectWise sold to private equity (Thoma Bravo, 2019). Kaseya, PE-backed since 2013, swallowed Datto for $6.2B. The tools you built your shop on, rolled up and milked.
SaaS
2011 →
The lift
Office 365 (2011) killed the on-prem Exchange server in the closet — everything moved to the cloud.
The toll
Then the sprawl: endless subscriptions and an attack surface no one imagined — and a whole industry selling you the tools to secure the software they already sold you.
AI
2022 → now
unfoldingThe lift
- 2022–23 It replaced the search box. Trained on a generation of human work, AI became how people find answers — almost overnight.
- Since It reshaped industry after industry. And it did it at a loss, subsidized to win the land grab. The prices aren't real yet.
- Now It writes the code. Coding agents are replacing traditional development. The tooling we run our businesses on is being rewritten in real time.
The toll
The bill is coming — metered, capped, marked up — and worst of all if you're locked into a platform taking a cut of everything you build.
This time, we're early.
This is the most fun the work has been in years — and the risk is just as obvious. AI is cheap because it's subsidized; the bill is coming, and it lands hardest if you're locked into a platform that marks up every token. For once we're at the scene before the crime — so we can build the alternative ourselves: open, model-agnostic, ours.