Choosing ChatGPT or Google AI for Coding
Compare ChatGPT and Google AI by coding harness, usage limits, Codex, Antigravity, Jules, and separate API billing.
Articles
Investigating Zenn, note, GitHub Pages, and Your Own Site
Compare canonical sources, URLs, discoverability, automation, and operations before choosing a personal technical-blog platform.
Choosing GPT-5.6 Sol, Terra, and Luna by Workload
Use official GPT-5.6 model data to compare prices, context, reasoning effort, and escalation criteria.
Reading LLM Usage and Pricing by Token Category
Separate input, cache, output, and reasoning usage, then recalculate an API bill from the recorded quantities.
Comparing OpenCode Go and Zen by Billing Path
Investigate OpenCode Go's flat allowance and Zen's usage balance through models, limits, fallback, and BYOK.
Investigating the OpenSpec OpsX Workflow
Treat OpenSpec explore, propose, apply, and archive as an artifact and review flow for AI-assisted development.
Separating GGUF Weights, Tokenizer, and Config in vLLM
Distinguish quantized weights, tokenizer, and Hugging Face configuration when starting GGUF models with vLLM.
Migrating vLLM Structured Outputs from Legacy Guided Fields
Move from guided_json to structured_outputs or response_format while keeping JSON Schema and semantic validation boundaries explicit.
Separating vLLM Token Limits and Reasoning Settings
Tune local inference by separating max-model-len, request limits, batching, GPU memory, and reasoning parsers.
Using 1Password CLI Safely in WSL
Separate the Windows and WSL boundary and use op run, op read, and TOTP without leaving secrets behind unnecessarily.
Separating the Roles of AGENTS.md, Skills, and MCP
Separate always-on project rules, reusable procedures, and external connections to reduce duplicated configuration and excessive permissions.
Troubleshooting TUI Keybindings in AI CLIs
Use Shift+Enter, Ctrl+J, Vim mode, and the status line to isolate terminal, shell, and TUI behavior.
How to Evaluate a New AI Coding Tool
Separate model and harness, then evaluate AI coding tools with reproducible tasks, quality, time, cost, and safety.
Separate the Model, Reasoning Effort, and Harness
Separate model capability, reasoning settings, harness behavior, and execution conditions when comparing Codex and OpenCode.
Starting a Personal Technical Blog with Minimal Astro
Build a small technical blog with current Astro content collections, dynamic routes, RSS, and static hosting.
Using Atuin Shell History by Search Scope
Separate Atuin's matching modes from history scopes and build a Ctrl-R workflow with workspace, directory, and session filters.
Safely Diagnose Cloudflare R2 and rclone Operations
Verify R2's S3-compatible setup and 403 meanings, then separate prefix copy, verification, and deletion.
Separating Codex Sessions, Resume, and Working Directories
Distinguish saved Codex sessions from shells, choose resume or fork, and diagnose contention without destroying history.
