Blog
Notes sent back from the road
Short, practical pieces on coding agents, models, fares and the protocols underneath.
English
Estimate Monthly AI Spend for a Two-Person Startup
Build a monthly AI budget from coding sessions, product requests, token usage, and a practical buffer in seven desk-ready steps.
English
One API Key Per Project for Predictable Agent Bills
A practical walkthrough for separating agent usage, controlling spend, and keeping RelayVano billing easier to forecast.
English
Use Cheap Models for Drafts and Strong Models for Finals
A practical workflow for drafting with lower-cost AI models and reserving stronger models for the final pass.
English
Prompt Caching: A Travel Itinerary Walkthrough
Learn prompt caching with a practical travel itinerary example: structure reusable context, test cache hits, and compare token costs.
English
Gemini 4 Argon: An Image-and-Text Prompt Walkthrough
Build clearer image-and-text prompts for Gemini 4 Argon with a practical walkthrough covering image labels, evidence, uncertainty, and review.
English
Hunyuan 4 and MiMo V2.6 Pro: A Practical Fit Test
A desk-ready walkthrough for testing Hunyuan 4 and MiMo V2.6 Pro on language quality, tool use, latency, and cost per accepted result.
English
GPT-5.6 Sol for Math: A Desk-Side Walkthrough
Follow a practical GPT-5.6 Sol workflow for math-heavy tasks: define constraints, derive equations, verify results, and save repeatable tests.
English
Claude Fable 5.1: Build Creative Writing Into Apps
Follow seven practical steps to prototype, test, and ship a Claude Fable 5.1 creative writing feature with clear controls and measurable quality.
English
Context Windows for a 200-File Repository
Measure your repository, budget tokens, and choose useful context window sizes with a practical seven-step walkthrough.
English
GLM-5.3 for Bilingual Support: A Desk Walkthrough
Build bilingual customer-support drafts with GLM-5.3 using a practical workflow for shared facts, terminology, review, and approval.
English
Grok 4.7: Quick Opinionated Summaries in 7 Steps
Turn long documents into concise, opinionated summaries with Grok 4.7 using a practical seven-step workflow and reusable prompts.
English
Kimi K3: A Chinese-English Technical Document Workflow
Follow a practical Kimi K3 workflow for long Chinese-English technical documents, with terminology controls, source references, and verification.
English
Claude Haiku 4.5 First, Bigger Model Second
Build a two-stage AI workflow: use Claude Haiku 4.5 to gather evidence, escalate difficult decisions, and measure whether the extra pass helps.
English
Onboard New Hires with Claude Code Codebase Tours
A practical, seven-step Claude Code walkthrough to help new hires map a repository, trace a feature, run tests, and prepare their first change.
English
Fix Front-End Accessibility with Coding Agents
Follow a practical, seven-step workflow to find accessibility barriers, guide a coding agent, and verify front-end fixes.
English
Remote Pair Programming with a Coding Agent over SSH
Follow a practical SSH workflow for pairing with a coding agent, keeping sessions alive, reviewing changes, and testing safely.
English
Read Foreign-Language Docs With a Coding Assistant
A practical walkthrough for reading documentation in another language, preserving technical details, and testing what your coding assistant explains.
English
Clean Up Legacy Shell Scripts with Codex CLI
A practical, seven-step walkthrough for auditing, fixing, testing, and safely deploying legacy shell scripts with Codex CLI.
English
Run a Coding Agent in a Docker Dev Container
Set up a Docker dev container, connect a coding agent, protect credentials, and review your first AI-assisted code change.
English
Gemini CLI: Summarize a Repo Before Code Review
Follow a practical Gemini CLI walkthrough to map a repository, inspect a pull request, and build an evidence-backed code review checklist.
English
Coding Agent Workflows for Notebook Data Scientists
A practical notebook workflow: scope agent tasks, protect data, extract tested functions, and verify results with clean execution.
English
Teach Your Coding Agent Your Team’s Naming Rules
A practical, numbered walkthrough for documenting naming conventions, guiding coding agents, and checking generated code before review.
English
Coding Agents for Safer Interface Localization
A practical walkthrough for translating UI strings with coding agents while protecting placeholders, plural rules, context, and repository scope.
English
Let Claude Code Test Legacy Code Without Rewriting It
A practical walkthrough for using Claude Code to capture legacy behavior, write reliable tests, and keep production changes out of scope.
English
Coding Agent Etiquette for Open-Source Contributions
Follow a practical walkthrough for respectful, reviewable AI-assisted contributions across programming languages and global open-source teams.
English
How a Coding Agent Reads a Monorepo: 8 Steps
Follow an eight-step workflow to help a coding agent navigate a monorepo, manage context limits, and verify changes without scanning everything.
English
Claude Desktop and RelayVano for Document Work
Connect Claude Desktop to a RelayVano-backed MCP tool, then review long documents with source references and controlled data sharing.
English
Gemini CLI First Steps Outside the United States
Install Gemini CLI, check regional access, run a safe first task, and understand authentication, billing, and relay configuration.
English
Pair Across Time Zones With an Agent Handover
A practical walkthrough for using a coding agent to preserve decisions, verify progress, and hand work to a teammate in another time zone.
English
Run Codex CLI Behind a Corporate Proxy While Travelling
A practical walkthrough for configuring Codex CLI, corporate proxies, trusted certificates and relay access while travelling.
English
Keep Claude Code Alive on Unreliable Hotel Wi-Fi
A practical walkthrough for protecting Claude Code sessions from hotel Wi-Fi drops, with checkpoints, tmux, safe reconnects, and recovery checks.
English
Set Up Claude Code With a Custom Base URL
Configure Claude Code with a custom API base URL on Windows, macOS and Linux, then test authentication and save your settings.
Or browse by topic: