News & guides
Batching Requests to Save Money: When It Pays
Batch endpoints trade latency for a large discount, and packing items into one prompt amortises overhead. Here is when each pays and when neither does.
· 8 min read
Best Model for Code Review: Optimise for Precision, Not Recall
An automated reviewer that flags everything gets muted within a week. How to pick and evaluate a model on false-positive rate, and why review has odd economics.
· 9 min read
Best Model for Debugging: Judge It on How It Handles Being Wrong
Debugging is a search problem, so the model that wins is the one that forms testable hypotheses and abandons them fast. How to evaluate that properly.
· 9 min read
Best Model for Test Generation: Coverage Is Not the Metric
Models are very good at producing tests that pass and prove nothing. Use mutation score to pick one, and treat test writing as your cheapest high-volume task.
· 9 min read
Building a Chatbot From Scratch: The Parts Nobody Mentions
The model call is twenty lines. The other ninety percent is conversation state, idempotency, abuse limits and knowing when a reply was wrong. A build order that works.
· 9 min read
Building Your First Agent: A Loop, Three Tools, One Guard
Skip the framework. An agent is a while loop over a tool-calling model. Here is the minimum that works, and the four things that break it first.
· 8 min read
Choosing an AI Provider: The Questions to Ask
A checklist of the questions that separate providers you can plan around from ones you cannot, covering model transparency, limits, compatibility and exit terms.
· 9 min readCline Setup Guide: Pointing It at a Custom Endpoint
Configure Cline against any OpenAI-compatible base URL — the provider fields, the model configuration block that people skip, and separate Plan and Act models.
· 8 min read
Context Window Cost Tradeoffs: Bigger Is Not Free
A million-token window changes what you can do and what you will pay. Here is the arithmetic for filling one, and when retrieval beats stuffing on cost.
· 8 min read
Context Windows: Why More Is Not Always Better
A million-token context window sounds like the end of chunking. In practice models get slower, pricier and less accurate long before you fill it. Here is why.
· 8 min readContinue.dev Setup: config.yaml for a Custom Endpoint
Point Continue at any OpenAI-compatible base URL using config.yaml — model roles, apiBase, request options, and why the autocomplete slot needs its own model.
· 8 min read
Cost Per Pull Request: A Unit Metric Worth Tracking
Total AI spend tells you nothing actionable. Cost per merged pull request ties inference to delivered work and exposes exactly where the money goes.
· 8 min readReady to try it?
Every model behind one key. Start free and upgrade when it sticks.