Blog

Building in public. Sharing what we learn.

Diagram of Chrome’s browser process connected to three tab renderers, an extension, and a GPU process, with each process’s work labeled

Why Does Chrome Use So Much Memory? Read Chrome Task Manager and Fix It

Why does Chrome use so much memory? Learn what its processes do, how to read Chrome Task Manager, and which fixes, like Memory Saver, free RAM.

Comparison table showing which code navigation approach fits each repository question, its setup needs, and its main limitation

Claude Code LSP: Choose a Navigation Tool for Your Agent

Claude Code LSP adds language-server navigation, but grep, semantic search, and Tree-sitter answer different questions. Compare their setup and limits.

A diagram comparing an agent's path through Chrome DevTools MCP and Playwright MCP to Chrome or multiple browser engines, including where browser profiles are stored

Chrome DevTools MCP vs Playwright MCP: Choose by Browser Job

Chrome DevTools MCP vs Playwright MCP: compare debugging tools, browser profiles, and measured context cost before you connect one to Claude Code or Codex.

Decision diagram mapping lid open or closed and AC power or battery to the relevant Mac sleep assertion and whether it stays awake

Caffeinate Mac Command: Keep It Awake for a Job or With the Lid Closed

Caffeinate Mac command guide: choose the right caffeinate flag for a build, download, display, or closed lid, then check and release the power assertion.

November 2026 Muvon release round: 105 releases across Octomind, Octolib, OctoHub, Octofs, Octoweb, Octocode, Octobrain, Synx and Yoda, with Unsame and mcp-variety

Release Round, November 2026: 105 Releases, Unsame, MCP Elicitation and Verified File Edits

105 releases across nine products. Unsame launched, Octoweb added AI chat history and calendar tasks, Octomind gained MCP elicitation, MCP OAuth, evidence checks and scheduled tools, and Octofs changed its remote MCP server authentication and edit contracts.

Diagram showing Claude in Chrome passing requests through its extension to logged-in browser tabs, with page content crossing a trust boundary back to Claude

Claude in Chrome Extension: How to Install It, Use It, and Control Browser Access

The Claude in Chrome extension can read and act on logged-in sites. See how to install it, connect Claude Code to Chrome, choose approvals, and limit risk.

Unsame Playbook Engine flow — an input goes to Unsame, which returns a protocol with situation, hooks, a different approach each time, form and voice, which the customer's own model turns into the final text

Introducing Unsame: Stop Repeated AI Writing Patterns in Every Post, Reply and Changelog

Unsame is a Playbook Engine that stops repeated AI writing patterns: a fresh, fitting approach before each generation. Remote MCP server, HTTP API, SDKs.

October 2026 Muvon release round: Octoweb 0.19, Vext 1.7.0, Timex 1.7.0 and TypeTab 1.2.0 with automatic updates, Octomind 0.60.1 with ChatGPT sign-in and Windows install, Octolib 0.45.1, OctoHub 0.13.0, Octofs 0.19.0, Octocode 0.27.0, Octobrain 0.14.8 and Synx 0.1.7

Release Round, October 2026: Four Mac Apps That Update Themselves, a Browser That Reads Mail, and an Agent on Windows

Eighty-eight releases across eleven products in three weeks. Octoweb, Vext, Timex and TypeTab moved to automatic updates; Octoweb 0.17–0.19 became a mail client; Vext 1.7 added Meeting Assist; Octomind 0.56–0.60 signs in with ChatGPT, installs on Windows and streams its subagents.

A comparison of MCP tool definitions, on-demand tool search, and code execution, with token measurement steps for Claude Code, Codex, and the Anthropic API

Code Execution with MCP: Measure Tool Tokens Before You Rewrite Your Agent

Code execution with MCP can shrink tool context, but first measure which tokens come from definitions and which come from results. Here is a Claude Code and Codex procedure, plus a decision table for direct calls, tool search, and code execution.

Worked monthly AI token cost estimate comparing OpenAI, Anthropic, and Google API models

AI Token Cost: Estimate Your Monthly LLM Bill

AI token cost is the price of each million input and output tokens. Estimate a monthly LLM bill from workload, model rates, cache use, and reasoning, then compare it with measured usage.

Terminal-style diagram showing Claude Code context sources, auto-compaction, and choices to compact, clear, delegate, or start a new session

Claude Code Context Window: What Fills It and When to Reset

Claude Code context window guide: see what fills it, read /context, choose /compact or /clear, and decide when a 1M window or fresh handoff helps.

A builder’s decision map connecting 22 of the best books for startups to product problems and actions for this week

Best Startup Books for Builders: What to Learn Before You Ship

The best startup books for builders and founders, with each book’s core idea, the common mistake it targets, and what to do instead on validation, positioning, pricing, growth, and quitting.