The Taku Blog
Ideas worth building on.
Stories, product notes, and perspectives on how AI gets discovered, used, remixed, and shared.
September 13, 2026
Git Workflow & Versioning Skill for AI Coding Agents
Addy Osmani's git workflow skill teaches an AI agent atomic commits, short branches, and semantic versioning. What it enforces and how to run it.
September 13, 2026
Home Energy Tune-Up: AI Skill to Cut Energy Bills
An AI skill that turns your bills and home details into a free-first, renter-aware energy plan with measurement steps. What it does and doesn't do.
September 13, 2026
HyperFrames Video Blueprint: Plan Before You Render
A planning skill adapted from HeyGen's open-source HyperFrames: timed beats, scene cards, and acceptance checks for a video, without rendering it.
September 13, 2026
Medal Forge SVG: Design a Medal or Badge as SVG
An AI skill adapted from Medal Forge: pick a shape, material, ribbon, symbol, and title, and get a deterministic 512×512 SVG medal. Runs locally.
September 13, 2026
Plain Language Skill: Explain Without Dumbing Down
An AI skill that rewrites dense technical material for a specific reader while keeping the numbers, conditions, and caveats that matter. How it works.
September 13, 2026
Prototype Skill: Throwaway Code for Design Questions
Matt Pocock's Prototype skill has your AI agent build disposable logic demos or UI variations to settle a design question. How it works and how to run it.
September 13, 2026
SaaS Metrics Health Check: MRR, Churn, CAC and LTV
An AI skill that turns raw SaaS numbers into MRR growth, churn, CAC, LTV, and NRR, labels each healthy, watch, or critical, and ranks what to fix first.
September 13, 2026
Savings Goal Timeline Review: AI Savings Planner
An AI skill that turns savings goals, dates, and balances into a monthly timeline and flags when goals don't fit. What it does and doesn't do.
September 13, 2026
Schema-Grounded SQL Queries: Plain English to SQL
An AI skill that reads your database schema and turns plain-English questions into commented SQL for PostgreSQL, BigQuery, MySQL, and more.
September 13, 2026
TikTok Content Planner: Scripts and Content Calendar
An AI skill that plans TikTok content pillars, hook-first scripts, and a review calendar without posting or touching your account. How it works.
September 13, 2026
Web App Testing Skill: Check Your App with Playwright
An agent skill adapted from Anthropic's webapp-testing: browser checks, screenshots, and console errors for your local app. How it works and how to run it.
September 13, 2026
Which Claude Model Is Best for Coding?
Anthropic points complex agentic coding at Opus 5 and everyday coding at Sonnet 5. Where Fable 5.1 fits, Claude Code's defaults, and the effort dial.
September 10, 2026
Copy Editing Skill: Tighten Marketing Copy You Have
A Claude skill that reviews and sharpens existing marketing copy — what it does, what's inside it, and how to run it without cloning a repo.
September 9, 2026
AI Prompt Marketplaces: Are Paid Prompts Worth It?
Prompt marketplaces sell text you could write yourself. Where they hold value, why most prompts depreciate, and what's worth paying for instead.
September 9, 2026
AI Source Finders: Tools That Cite, and How to Check
AI source finders answer with citations. Which tools really retrieve sources, the failure mode that matters, and a one-minute check.
September 9, 2026
AI Video Finders: Searching Inside Video
AI video search finds the moment, not the file. How transcript, visual, and semantic search differ, what each misses, and how to pick.
September 9, 2026
Using Claude Code With Cursor: Run Both Together
Claude Code and Cursor aren't an either/or. How to run the terminal agent alongside the editor, which does what, and the conflicts to avoid.
September 9, 2026
Claude Code Use Cases: What It's Genuinely Good At
Ten Claude Code use cases sorted by how reliably they work, the property they share, and the tasks worth keeping away from an agent.
September 9, 2026
Claude Code vs Claude Desktop: Chat, Cowork, Code
Claude Desktop now has Chat, Cowork and Code tabs. What each can reach, how approvals work, and when Claude Code in the terminal is still the better fit.
September 9, 2026
How to Update Claude Code (and Why It May Not)
Claude Code auto-updates on native installs but not on Homebrew, WinGet, or apt. How to update each, pick a release channel, and pin a version.
September 9, 2026
Is Claude Code Worth It? Who It Actually Suits
Claude Code needs a paid plan and a terminal. Where it genuinely pays off, where it doesn't, and the four questions that answer this for you.
September 9, 2026
Manus AI Review: What It Does Well, Who Should Skip
An honest look at Manus — the task-completion shape, where it genuinely delivers, the limits inherent to general agents, and who it doesn't suit.
September 9, 2026
How to Use Manus AI: What It Does and How to Brief It
Manus takes a goal and produces a finished artifact — slides, a site, a design. How to brief it well, where it fits, and how to reuse what worked.
September 7, 2026
Claude Agent SDK: Which Anthropic SDK You Need
The Claude Agent SDK is Claude Code as a library. How it differs from the Anthropic Client SDK, why it was renamed, and which one you need.
