Show Notes
One CTO spent two weeks tearing frontier models out and ended up back where they started. Another spent a few days building a tested escape route for what actually mattered.
Build your own CTO Playbook at www.theCTOplaybook.com - the leadership platform built for the full CTO journey. Coaching, podcast, and community to help you lead with clarity, confidence, and strategic impact.
The safe CTOs aren't the ones who trust their AI provider, and they aren't the ones tearing frontier models out of every workflow; they're the ones who've triaged their dependencies and tested a real escape route for the handful that matter. I'm walking through this one solo because it's a structural risk most engineering leaders felt this year but haven't priced. Frontier AI is a supplier you can't audit, and it can be switched off by forces sitting outside your contract. Your invoices can be paid in full, and the capability can still vanish.
Most of us onboarded this dependency without ever writing it down. It spread the way water finds its level, and it became load-bearing before anyone noticed. Where in your stack would a frontier model going dark tomorrow actually hurt, not annoy you, but hurt?
The third option is triage: trust the dependency where stakes are low, temper it where you're more reliant than is wise, and override it with a tested alternative for the workflows that would cost you the business.
Two CTOs I work with responded to the same news in almost opposite ways, and only one of them still has a working escape route.
You'll Learn:
[0:00] Introduction [2:06] What actually happened when two frontier AI models went dark for eighteen days [4:26] Why a frontier AI model is a supplier you can't audit or control [6:37] The dependency most teams priced at zero and never wrote down [8:41] The fire drill that went off for someone else's building [11:39] Two CTOs face the same news and respond in almost opposite ways [15:02] Where in your stack would a model going dark tomorrow actually hurt? [19:27] Why an untested escape route is just a hope, not a fallback
Find more from Adam on LinkedIn and YouTube, and explore coaching, cohorts, and how you can stay up to date at theCTOplaybook.com, helping you build your own playbook for your path at your pace.
More from The CTO Playbook
Most leaders treat impostor syndrome as something they need to overcome. But what if the belief that you're not good enough is actually serving a purpose?
AI has made building software dramatically faster. The problem is that the rest of the engineering system hasn’t caught up. AI agents can now produce code, tests, documentation, migrations, and pull requests at a pace most engineering organisations weren’t designed to handle. But when everything around code review accelerates, review becomes the bottleneck. Pull requests pile up, reviewers spend more time reconstructing intent, and experienced engineers can end up reduced to repeatedly saying yes or sending work back for fixes.
AI adoption is getting easier for everyone. The harder question is whether you're using it to optimise the game you're already in, or to build a position that changes the game.
AI isn't just changing how engineers work. It's forcing technology leaders to rethink what makes management valuable.
Quantum computing may not be breaking your security today, but the decisions you make now could determine how exposed your organization is when it does. Conor Deegan explains why CTOs need to start preparing before Q Day arrives.