— ਲੇਖ
ਇਸ ਵਿਸ਼ੇ ਉੱਤੇ ਫੀਲਡ ਨੋਟਸ: AI
ਇਸ ਵਿਸ਼ੇ ਉੱਤੇ SetKernel Digital Inc. ਦੇ ਸਾਰੇ ਲੇਖ, ਨਵੀਨਤਮ ਪਹਿਲਾਂ।
-
MCP Servers in Production: When to Build, When to Skip
Model Context Protocol is the best primitive Anthropic shipped in 2024 — and the most over-applied. A field guide to when an MCP server is the right answer, when it's overkill, and how to operate one in production on Cloudflare Workers.
-
Structured output from language models: getting reliable JSON instead of prose
Free-text model output breaks pipelines. Here is how schema-constrained generation, tool calling, validation, and retries turn a language model into a component you can build on — and where the approach still fails.
-
RAG evaluation: how to know your AI answers are actually correct
A retrieval system that sounds confident is not the same as one that is right. How to build an evaluation set, measure retrieval and faithfulness, run human spot-checks, and catch regressions before your users do.
-
When Not to Use an LLM
A working checklist for keeping large language models out of the places they don't belong: deterministic logic, unverifiable outputs, high-stakes decisions, and problems a database query already solves.
-
Making Your Website Agent-Ready: What Actually Matters in 2026
AI agents and answer engines read your site before people do. What makes a site legible to them — structured data, markdown, an AI-crawler policy, and MCP.
-
Workers AI vs OpenAI: A Cost-Quality Matrix at Low Volume
Most production AI features default to OpenAI by reflex. They shouldn't. A real comparison of Workers AI, OpenAI, and Anthropic Claude across the four tasks that actually show up in client engagements — embeddings, generation, transcription, classification — with real numbers and a routing strategy that uses each for what it's good at.
-
Production RAG on Cloudflare Without LangChain
A 200-line RAG pipeline on Cloudflare Workers + Vectorize + D1 — the five primitives that matter, why frameworks rot at the wrong layer, and how to keep retrieval debuggable in production.