(Tag)
AI in products: LLMs, RAG, agents, cost, risks and real business value.
5 min read
The risks LLM applications face, why prompt injection cannot simply be forbidden and how to protect a product architecturally: least privilege, action checks and human confirmation.
5 min read
When an LLM support bot pays off, how it works — knowledge base, tools, human handoff — and how to launch it in stages without putting customers at risk.
5 min read
Where AI assistants really speed up development, what research says, the risks they bring and how to build them into your process without losing quality.
5 min read
How AI assistants choose sources, how to open your site to their crawlers and how to write text that is easy to quote. Practical techniques and technical signals.
5 min read
What an AI feature really costs, how to estimate token spend with a formula, how to cut it and when a self-hosted model beats a cloud API.
5 min read
RAG lets an assistant answer from your knowledge base without fine-tuning. We cover the components, a Python example with pgvector and why answers go wrong.
6 min read
Calling a model takes two lines of code; the value of an AI feature lies in context, validation, limits and logging. Here is the architecture and the post-launch pitfalls.