Insights
We are the sober practitioners in a room full of breathless futurists.
Framework 01
Why some organizations get better with AI over time. Most don't.
The organizations pulling ahead are not the ones with the best models. They are the ones whose flywheel is spinning. Most organizations have all four components broken and spinning independently, so the flywheel never engages.
“AI doesn't get better with time. Organizations that invest in the flywheel do.”
Framework 02
A layered diagnostic for why AI investments underperform.
Most organizations diagnose at Layer 3 or 4. The actual root causes are almost always at Layers 0 to 2. The instinct is to upgrade the model or buy a different tool. The actual problem is how the organization operates around the technology.
“When your AI investment underperforms, the fix is a leadership decision, not a vendor decision.”
Framework 03
A 5-level progression for organizations, not individuals.
Every executive reading this can place their organization. The gap from Level 2 to Level 5 is where our engagements happen.
“Where does your organization actually sit?”
Framework 04
How regulated industry leaders should think about the EU AI Act.
Companies that built GDPR compliance-first didn't just survive the regulation. They built defensible data infrastructure competitors couldn't replicate. The same opportunity exists now with the EU AI Act. The window to move first is closing.
“Every company sees the EU AI Act as a compliance burden. The ones that will win see it as a moat.”
| Compliance Burden | Compliance as Moat |
|---|---|
| Cost center | Valuation premium |
| Risk to minimize | Capability to build |
| Deadline to survive | Standard to set |
| Legal problem | Product strategy |
| Audit anxiety | Enterprise sales advantage |
Essays
Our founder's personal site hosts the thesis essays that informed the way we work. Read them there.
The $684B spent on AI in 2025. The 80% that didn't deliver. Why the gap is not technology, and what to do about it.
Read on golubovski.com →EssayCompliance-as-moat. The companies that built GDPR-first didn't just survive, they built defensible data infrastructure. The same opportunity is closing now.
Read on golubovski.com →EssayThe 2020 scaling playbook. Why it stopped working. What replaces it. Written for CPOs and founders who know this is the year.
Read on golubovski.com →Downloadable resources
The flywheel that separates organizations pulling ahead from everyone else. Better Specs → Better AI Outputs → Better Eval Data → Better System Design.
Download the framework →A 30-question self-check on where your operating model is breaking. Spec discipline, eval practice, governance, team structure.
Download the checklist →And what to do about it.
