Guides for production LLM systems

Implementation patterns, technical comparisons, and decision rules for engineers building with language models and agents.

All guides

Deep-teal field with three rises and a precise lower notch

·AI Agents

Treat Reasoning Controls Like API Migrations

Ship the model, reasoning mode, and history adapter as one versioned protocol.

A vermilion field with progressively deeper rule removals

·AI Agents

Delete System-Prompt Rules One Regression at a Time

Tie every prompt workaround to a failure case, then test its removal when the model changes.

Two unequal cobalt lobes joined through a narrow waist

·AI Agents

Your Agent Memory Needs Two Clocks

A single timestamp cannot distinguish a late correction from a real-time state change.

Two ochre fields separated by an unbroken ivory boundary

·AI Agents

The Tool List Is Not an Authorization Boundary

Hiding a tool from an agent changes its options, not the authority behind them.

Aubergine field with four progressively deeper recesses

·AI Agents

What Pass Rates Hide About Long-Horizon Agents

Milestone reach rates reveal the transition where agent reliability collapses.

Two olive cache paths with unequal lengths

·LLM Infrastructure

Sliding-Window Prefix Caches Need Physical Proof

Hybrid attention turns prefix caching into a memory-consistency problem across KV pools with different lifetimes.

Two offset vermilion plates separated by a commit channel

·AI Agents

Speculative Agents Need a Commit Boundary

Speculative execution cuts agent latency only when prediction and commitment are separate operations.

A deep-teal form folding into two parallel arms

·AI Agents

Loop / Harness Engineering

Agent loops turn prompts into one part of a production harness: tools, state, checks, approvals, and exit conditions.

Ochre form divided by two nested ivory incisions

·AI Agents

Plain-Text Context Files for LLMs

Plain-text context files are the git-native, zero-ops way to steer AI agents.

Three cobalt inlet lobes merging into one body

·AI Agents

How to Build an MCP Server

A guide to building an MCP server for Markdown note taking

Three fused olive chambers separated by alternating notches

·AI Agents

Structuring tools for AI Agents

A short conceptual guide to organising tools in agentic workflows

Irregular and smooth aubergine forms divided by a clean cut

·LLM Infrastructure

The Modern Document Processing Stack

A simple engine for converting any file type to Markdown.

Three deep-teal forms fanning from a shared origin

·LLM Infrastructure

Decoding Methods for LLMs

A verbose explanation of common auto-regressive decoding methods: temperature, top k and top p.

Two vermilion forms facing across a measured gap

·Structured Generation

Structured Generation Shoot-out

A comparison of some of the most popular libraries that constrain LLM generation.

Compare open-source AI frameworks

Filter projects by language, age, funding, size, and GitHub stars.

Open the leaderboard