FundamentalsPositional Encoding: How a Transformer Knows Token Order
Attention is order-blind by default. How position gets injected, why the method decides how far context can stretch, and what you see when it fails.
ReadNotes on models, agents, and what inference costs
Practical writing for developers building with large language models — how they work, how to pick one, and how to keep the bill predictable.
Fundamentals
GuidesPre-Commit Hooks That Call an LLM: Worth It or Not?
How to wire a model into a git pre-commit hook, which checks actually belong there, and the latency and cost budget that decides whether the team keeps it.
Read
Cost & PricingPrepaid vs Postpaid AI Billing: Credits or an Invoice
Prepaid credits cap your downside and expire. Postpaid invoices never block work and arrive after the damage. How to pick when usage is unpredictable.
Read
FundamentalsPretraining vs Post-Training: Where a Model Gets Its Behaviour
Pretraining decides what a model knows. Post-training decides how it behaves. Knowing which stage owns a problem tells you whether prompting can fix it.
ReadCost & PricingPrompt Caching Savings: The Arithmetic Nobody Runs
Cached input is priced far below uncached input. Working out what that is worth for your workload, and what prompt structure it demands.
Read
AI AgentsPrompt Regression Testing: A CI Gate for Fuzzy Output
Prompt edits regress silently and every score is noisy. How to build a golden set, pick a scorer, handle variance and gate merges without a permanently red build.
Read
GuidesPrompt Templates and Versioning: Treating Prompts as Code
Prompts drift, break silently and get edited in production. How to template them safely, version them properly and know which version produced which output.
Read
GuidesProxying LLM Traffic: What a Gateway Has to Get Right
Putting a gateway in front of an inference provider centralises keys, budgets and routing. Here is what breaks if you get streaming, headers or cancellation wrong.
Read
Cost & PricingQuantization Cost Savings and the Quality You Trade Away
Lower precision cuts memory and raises throughput, which is what makes single-GPU hosting possible. What it costs in quality is not evenly distributed.
Read
ModelsQwen 3.5 Coder vs DeepSeek V4 Pro: Specialist or Generalist
A code-tuned model against a large general MoE. What each design buys you, where the published numbers stop helping, and how to decide on your own repo.
Read
ModelsQwen 3.6 27B vs Max: Shipped Model or Preview Tier
One is a dense 27B you can run on a single GPU today. The other is a preview whose numbers can move. How to pick a tier inside one model family.
ReadModelsQwen 3.6 Guide: The Frontier Model That Fits on One GPU
Qwen 3.6 27B is dense, not mixture-of-experts, and runs on a single GPU while scoring 77.2 percent. Why that combination matters more than its position on a leaderboard.
ReadShowing 205–216 of 404 articles