Agents should not call tools
Most agent stacks let the model pick a tool and log the call. I think that is the wrong default. Agents should ask for business capabilities instead.
Venkatesan Ramachandran
Progressive Autonomy for AI Agents: From Intern to Principal
Treat AI agents like new staff. They start with read-only access and earn more over time. Someone watches for odd behaviour, and there is a kill switch.
Why Your Company Needs One API Platform, Not Six
We used six separate tools to run our APIs. Here is why we moved to one platform, what we got wrong on the way, and when one platform is the wrong choice.
Multi-Tenant SaaS with PostgreSQL Row-Level Security: Beyond the Basics
How we kept tenants apart in a Go SaaS platform with PostgreSQL row-level security. The things that bit us, and the point where RLS stopped being enough.
The Block Pipeline Pattern: Composable Middleware Without the Mess
Middleware chains get messy as a product grows. A small block interface keeps an API gateway readable, cheap to extend and easy to test. Plus my mistakes.
Last Year: Invest in AI. This Year: Cut Costs.
Boards now want AI investment and a smaller tech budget at the same time. How to decide what to fund, what to stop, and how to defend an AI platform budget.
Why Every Enterprise Needs an AI Gateway
Most companies send LLM traffic straight from apps to providers. An AI gateway adds DLP, prompt-injection checks, cost caps, routing and an audit log.