Field notes
Notes from the work.
What we are learning about AI, from choosing models to building systems people can rely on.
RAG vs Agentic Retrieval: What Actually Ships in Production
A practical guide for CTOs and product leaders comparing classic RAG pipelines with agentic retrieval loops, their trade-offs and the hybrid systems that ship.
From Chatbots to Agents: When AI Starts Doing the Work
Chatbots answer questions. Agents use tools to execute workflows. The operating model, safeguards and architecture needed to ship agentic AI safely.
The Last Mile Problem: Why AI Coding Agents Don’t Ship Software
AI coding agents generate code quickly, but shipping the right feature still depends on scope alignment, evidence and verification.
Fine-Tuning vs. RAG: Choosing the Right Architecture
Should you fine-tune a model on your data or connect it to your data with RAG? A guide to the trade-offs, costs and production patterns.
The Multi-Model Strategy: Why You Shouldn't Lock In to OpenAI
How multi-model routing can reduce cost, improve reliability and support data privacy in enterprise AI systems.
Beyond ‘Vibe Checking’: How to Evaluate AI Systems at Scale
How to run evaluations like tests, use LLM-as-a-judge safely and prevent regressions when models and prompts change.
Understanding Prompt Caching: Reduce AI Costs and Latency
How prompt caching works in modern LLM APIs, including cache prefixes, TTL, cost savings and when to use it.
Why Embeddings Matter: A Technical Guide for Business Leaders
From vectors to practical applications: how embeddings enable semantic search, recommendations and retrieval-augmented generation.
Writing Effective Prompts for LLMs
Practical techniques for system prompts, examples, reasoning tasks and structured outputs across modern language models.
Understanding Generative Search and Its Application in Your Company
How generative search combines vector databases with language models to provide accurate, current responses.
Demystifying Transformer Models in Machine Learning
The architecture powering modern AI: tokenisation, embeddings, attention mechanisms and why transformers matter to applied AI.