Notes on data, AI, IT
and security
No marketing fog. The way I think about real problems with founders and managers.
Event-driven architecture: what managers need to know before committing
Events and message queues solve real coordination problems between services. They also introduce complexity that is easy to underestimate from a project plan.
The gap between experiment and production: why ML models never reach work
Most ML projects show good results in experiments and perform poorly in production. I look at why this happens.
When a good model goes bad: drift, detection, and business cost
A model that passed every test at launch can quietly degrade over months. Understanding why helps you decide how much monitoring is worth the investment.
Real-time analytics: when it works and when it is expensive theatre
Streaming data and real-time dashboards have become a fashionable requirement. I look at when this actually solves a real problem.
Where RPA breaks: the process complexity ceiling
Robotic process automation delivers fast wins on simple, stable workflows. When processes have exceptions, it starts to cost more than it saves.
Warehouse automation: what is behind pick rate numbers
How to think through the economics of warehouse automation and why picking speed is not the main metric.
Cloud vendor lock-in: the tradeoffs that actually matter
Lock-in is not automatically bad. The question is whether you are trading flexibility for something you actually need, or just for a lower list price today.
The unexpected cloud bill: what companies discover a year after migration
Why cloud costs end up higher than projected - and what you can do before it becomes a problem.
Colonial Pipeline: when a cyberattack stops physical infrastructure
The May 2021 Colonial Pipeline attack showed that the boundary between IT security and operational security has disappeared.
Data mesh: what it means for managers, not engineers
The data mesh concept is gaining popularity. I break down what it actually means and when it makes sense for a real business.
GPT-3 one year on: what actually changed for business
A year after GPT-3's release is a good moment to separate real shifts from noise and understand where language models already work.
API-first or integration spaghetti: the choice your architecture makes
Why the way your company's systems talk to each other determines growth speed and the cost of change.