Notes on data, AI, IT
and security
No marketing fog. The way I think about real problems with founders and managers.
Transformer moves beyond NLP: what it means
The transformer architecture that reshaped text processing is beginning to work with images and structured data. What this means for business.
API as a product: why internal integrations fall apart
How the approach to internal APIs affects the reliability of the whole architecture, and why an internal client deserves no less respect than an external one.
Feature flags in production: the promise and the maintenance debt
Feature flags let you ship safely and experiment cheaply - but every flag you add is a piece of logic you have to eventually remove.
Data catalog: what it gives you and when you can skip it
A practical look at a data registry: what it delivers, when it justifies the investment, and when it is premature complexity.
Hafnium and Exchange: the patch that waited too long
The March 2021 Microsoft Exchange mass exploitation showed that patch management is not a technical task - it is an organisational one.
MLOps: the gap between experiment and production
Why most ML experiments never reach production, and what to do about it at the organisational level.
SolarWinds: a supply chain attack explained for managers
What happened with SolarWinds, and why this incident changes the security conversation for companies that do not think of themselves as targets.
Kubernetes for small business: when orchestration is overkill
Why complex infrastructure often slows small teams down, and how to match the level of complexity to the actual task.
When a data lake turns into a swamp
Why data lakes often stop working one or two years after launch, and what to do about it.
CLIP and multimodality: the arrival of zero-shot behaviour
What OpenAI's CLIP release means for companies thinking about practical AI beyond text.
SolarWinds: supply chain risk now belongs in the risk model
The SolarWinds attack in December 2020 showed that trusted software can be an attack vector. What this means for how companies must think about their software suppliers.
A year-end data audit: five questions
The end of the year is a good time to check what state your company's data is in and what to fix before the new planning cycle begins. Five concrete questions for a leader.