How to choose a model
A task-shape-first method for picking a model tier: define the shape, set a capability floor and a cost-per-completed-task ceiling, benchmark on a private eval, then step down from frontier until quality breaks.
Guides
Practical writing on choosing models, controlling cost, and reading the field accurately.
Data checked 2026-09-06Understanding token pricing
How LLM API billing actually works: the four input meters, why resent conversation history dominates the bill, and why cost per completed task is the only number that ranks options correctly.
Context windows in practice
How token windows actually behave under load: why input and output limits are separate, why advertised length overstates useful length, what context costs, and which management strategy to reach for first.
How to read AI benchmarks
A working guide to benchmark families, contamination, saturation, harness effects, and the statistics that decide whether a leaderboard gap is real — plus a protocol for evaluating models on your own task.
Cutting API costs
A measurement-first playbook for lowering LLM API spend: instrument usage, take the free wins (caching, prefix hygiene, batch, retry hygiene) in order, then trade quality only against an eval.
Open weights or a hosted API
A decision guide for choosing between a hosted model API, self-hosted open weights, and open weights on a serverless platform — built around utilisation, licence terms, and compliance rather than headline price.
Retrieval, long context, or fine-tuning
How to choose among retrieval, long context, fine-tuning, and tool access — what each actually solves, where each fails, and the default path.
Tool use and agents
How the tool-calling loop actually works, when an agent beats a workflow, and the design, error-handling, gating, and observability decisions that decide whether the loop survives production.
Evaluating LLM applications
A working manual for building eval sets, choosing graders, running LLM judges, splitting train/test, and measuring cost, latency, and regressions on model upgrades.
Data privacy and compliance
How to review an AI vendor's data practices: what training vs. retention actually means, where consumer and API terms diverge, what zero data retention costs you, and the residency, subprocessor, redaction, logging, and regulatory questions to settle before you ship.
Your first API call
The request shape every LLM API shares, working examples for the major providers, and the four things that break in production.