Julien Vantyghem CTO of Quintess AI, automating industrial maintenance workflows.
On a Friday afternoon, I received technical details on a customer’s in-house authentication system, which enabled us to build federation. The team was maxed out and we needed to go fast, so I took it up. I fired up Claude Code and had a working solution in less than an hour. PR submitted for review, mission accomplished.
On Monday, one of our developers asked to chat because they had a similar problem and we could probably extract a reusable solution pattern. We hopped on a call, and as I started presenting my work to the developer, I realized I couldn’t clearly express what was there, why it was there, and I didn’t have clear insights on how to apply it to other cases. I recognized with an awkward laugh that it was vibe coded quickly and adjourned the meeting. Time wasted, poor impression on my teammate.
Cognitive Debt Is Putting Your Startup At Risk
On that Friday, I fell into the cosy arms of frontier models that not only do artifact generation but also understanding and decision-making for you. At no point in the process did I do any System 2 thinking, I just used muscle memory to vaguely steer the agent’s process toward a plausible solution and never acquired a firm grasp on what I was tackling. I borrowed cognition from the AI latent space and was now responsible for something I did not understand. I had incurred cognitive debt.
Accumulating cognitive debt happens naturally, because:
• The process feels right: Agent interfaces give you the illusion of understanding and decision. You get a dopamine shot when closing a task, irrespective of your actual cognitive gain.
• The result feels right: The polished aspect of AI-generated artifacts makes you trust them, especially if you’re venturing far from your area of expertise.
• The world tells you it’s okay: Output-oriented culture and AI hype don’t value human cognition and assume AI will handle it all eventually.
This phenomenon is not unique to software development. Your startup is likely incurring cognitive debt on important work like writing policy, contracts or product vision.
But there’s a problem with artifacts produced with cognitive debt. They’re not assets because you can’t stand behind, leverage or evolve them.
They’re liabilities that hurt you on leadership, efficiency and future velocity until you spend human tokens to internalize them. If you don’t, your stakeholders’ trust in them and you as their author erodes, and their value drops to zero.
Indulge in cognitive borrowing long and hard enough, and you end up owning an incoherent and complicated mass of stuff that you can’t understand or control. For individual teammates, it means anxiety and indifference to quality. For your startup, it means failure.
Knowing that risk, how can we safely leverage GenAI’s indispensable productivity gains?
Focus On Your Startup’s Theory
Margaret-Anne Storey coined the term “shared theory“ for software systems, as the collective understanding your team has about how the system works and why decisions were made.
I think this powerful concept is applicable to entire organizations.
Your startup’s theory is the understanding of the problems you have decided to tackle and how you are solving them. It’s your startup’s answer to: “What do we believe in?” “What have we learned so far?” “What have we solved and how?” and “What hypotheses are we testing and why?”
Your shared theory is your backbone; it guides all meaningful decisions. It grows with learning, shrinks with attrition and weakens with shortcuts.
Traditionally, your theory was concentrated in the head of the C-suite and partially conveyed through requirements to the team for execution. But with all team members now having unlimited execution capacity, this no longer holds. Realizing your company’s vision 10 times faster requires making 10 times more right decisions and having 10 times more discipline in maintaining coherence.
Here is the mindset shift I believe allows this controlled acceleration: Your startup’s theory is your most important asset, and it’s the whole team’s responsibility. All meaningful work takes it as input and contributes back to it. This applies across organizational units, departments and activities.
The main objective here is economic success, but I’d argue it’s also a matter of ethics. Ensuring a strong cognitive posture behind anything you provide internally or externally is respecting what collaboration and business mean in the first place. You owe it to your stakeholders.
Grow The Theory
Start by bringing your team together. The time AI hands back on execution is an opportunity to rebuild cognitive cohesion and foster cognitive ownership. Use it to talk through key problems and decisions as a group, break down silos, and make live design, share-out and deep review sessions a regular habit.
Alongside that, keep a living documentation system (your GTM strategy, team principles, product decisions, ADRs) written by humans, kept short and genuinely readable, with every evolution reviewed by the team before it’s committed. This is also the core context your internal agents should consume.
Protect The Theory
Start with transparency; discuss openly and harmonize how your team uses AI.
Add cognitive debt to the team’s shared vocabulary; everyone should understand what it is, what causes it and what it costs.
Protect human cognition deliberately: When a large task begins, scope it out without AI. When it’s finished, present the learnings and decisions without AI. These two human hours, instead of five AI minutes, are time well invested on your important cognitive checkpoints.
The Bottom Line
Left unchecked, cognitive debt accumulates quietly, one ungrounded decision at a time, until you’re leading a company you no longer fully understand.
The fix isn’t to use AI less. It’s to be deliberate about where your own thinking still has to happen. Treat your startup’s theory as a living asset that every interaction either strengthens or erodes. Ask before delegating to AI: Am I borrowing execution, or am I borrowing cognition?
The teams that prove most successful in the long term will be the ones who stay the rightful owners of important decisions.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
