News & guides
Cutting Token Usage Without Making Answers Worse
Most prompts carry substantial waste. Nine techniques that reduce spend while usually improving output quality, ordered by how much they actually save.
· 8 min read
Debugging LLM API Errors, Status Code by Status Code
A field guide to the errors an LLM API actually returns: what each status means, which ones are worth retrying, and how to reproduce the failure in one curl command.
· 9 min read
DeepSeek V4 Pro: Frontier Reasoning at Commodity Prices
A 1.6T MoE that activates 49B per token, ships under MIT, and costs $0.435 per million input tokens. Where V4 Pro is strong, where it is not, and Pro versus Flash.
· 10 min read
Embeddings Explained: Vectors, Similarity and Where It Breaks
Embeddings turn text into vectors so you can search by meaning instead of keywords. Here is how they work, what similarity really measures, and the failure modes to expect.
· 9 min read
Evaluating Agent Reliability: Past the Single Pass Rate
An agent that works four times in five is not 80 percent useful. Here is how to measure agent reliability in a way that predicts production behaviour.
· 10 min read
Fine-Tuning vs Prompting vs RAG: How to Adapt a Model
Three ways to make a general model do your specific job, with very different costs. Here is a decision rule based on what each one can and cannot actually change.
· 9 min read
Forecasting AI Spend Without Guessing
Most AI budget forecasts are a headcount multiplied by a hopeful number. Here is a model that decomposes spend into drivers you can actually measure and control.
· 8 min readFree AI API Tiers: How to Compare Limits That Keep Moving
Free tier numbers rot within weeks, and several vendors have stopped publishing them entirely. Here is a method for comparing them that survives the churn.
· 8 min read
Generating Commit Messages That Are Actually Useful
A diff shows what changed, never why. Where generation genuinely helps, a working prepare-commit-msg hook, and the Conventional Commits rules a model gets wrong.
· 8 min read
Generating Documentation With LLMs That Is Worth Reading
Most generated docs restate the function signature in English. What to generate instead, which formats tools can actually consume, and how to stop docs drifting from code.
· 9 min read
Generating SQL Safely: Least Privilege Beats Better Prompts
Text-to-SQL fails two ways: destructive queries and quietly wrong answers. Read-only roles and timeouts handle the first. The second needs a different kind of guardrail.
· 9 min readGenerating Tests With LLMs: Coverage Is Not the Goal
Models are very good at writing tests that pass and prove nothing. How to get suites that actually catch regressions, and how mutation testing tells you which ones do.
· 9 min readReady to try it?
Every model behind one key. Start free and upgrade when it sticks.