金 OKANE LAND

The Primer

Vendor-neutral, written to be useful first. How to actually use the AI tools you work with, from coding to content to automation.

A coral woodblock starfish priming a board with a brush beside a can of primer, the mascot of The Primer.
A hand-inked terminal window running the OpenCode TUI: a model-picker badge in the corner, a prompt, a proposed edit with a diff marked applied, and a status line reading any provider, bring your own key, open source; the Okane Land seal top-left and the starfish at the lower left.

Tools & Stacks

How to use OpenCode: the open-source terminal agent

OpenCode is a free, open-source AI coding agent that runs in your terminal and points at almost any model, 75+ providers or a local one on your own machine. It is the model-agnostic, MIT-licensed answer to a terminal agent, and the two things worth knowing first are which OpenCode is the real one and how you connect a model. This is the whole thing, install to first edit, and what it does and does not cost.

akira · 5 min

 A hand-inked cloud-IDE window: an activity rail, the Replit Agent building an app through a checkpoint list, and a live preview beside a Deploy button, with the Okane Land seal top-left and the starfish watching from the lower left.

Tools & Stacks

How to use Replit Agent: from prompt to deployed app

Replit Agent is an AI that builds a working app from a plain-language prompt inside Replit's browser IDE, then deploys it on the same page, with nothing installed on your machine. The part worth understanding before you start is the meter: Agent is billed by effort, one charge per checkpoint, and Replit does not publish a fixed price per checkpoint. This is the whole thing, first prompt to live URL, and how the billing actually works.

akira · 6 min

 A hand-inked editor window with code on the left and the Cline agent panel on the right, a Plan and Act toggle above a proposed diff with Approve and Reject buttons, the Okane Land seal top-left and the starfish reviewing from the lower left.

Tools & Stacks

How to use Cline: the open-source agent in your editor

Cline is an open-source coding agent that lives in your editor and terminal, and its whole design is that it proposes every edit and waits for you to approve it. Two things make it worth learning: it plans before it acts, and it connects to almost any model, including a local one at zero per-token cost. This is the whole thing, install to first approved edit, and the three ways to pay for the model behind it.

akira · 5 min

 A hand-inked browser window on v0.dev turning a prompt into a polished generated UI component, the Okane Land seal top-left and the starfish reading the result from the lower left.

Tools & Stacks

How to use v0: generate the UI, get real code

v0 is Vercel's AI builder: describe a UI or an app, and it writes real React and Next.js code you can preview, own, and deploy. Two things trip people up before they start, and both are here. It moved from v0.dev to v0.app, and it bills in credits metered on tokens, with a free tier capped at seven messages a day. This is the whole thing, first prompt to shipped, and what the credits actually buy.

akira · 6 min

 A hand-inked browser window building an app from a prompt, a token meter beside it showing a daily free allotment ticking down, the Okane Land starfish typing the prompt.

Tools & Stacks

How to use Bolt.new: build an app from a prompt

Bolt.new turns a prompt into a running full-stack app in your browser, with no local setup, because StackBlitz runs the whole dev environment in a browser tab. The thing to understand before you start is the meter: Bolt bills by tokens, with a real free daily allotment. Here is the whole thing, first build to owning the code, and how to make those tokens last.

akira · 6 min

 A hand-inked Zed editor window with code and an AI agent panel open on the right, a local model plugged in via a cable labelled Ollama, the Okane Land starfish at the keyboard.

Tools & Stacks

How to use Zed: the fast, open-source editor with AI

Zed is a code editor written from scratch in Rust for speed, open source, and free. Its AI is what makes it a vibe-coder tool: an agent panel that edits your code, a one-line path to a free local model, and the ability to run Claude, Codex, or Gemini CLI inside the editor over an open protocol. Here is the whole thing, install to agent.

akira · 5 min

 A hand-inked desktop window of LM Studio with a downloaded model and a chat, a local server toggle labelled port 1234 wiring out to VS Code, Codex, and Claude Code, the Okane Land starfish clicking start.

Tools & Stacks

How to use LM Studio: run local LLMs with a GUI

LM Studio is the friendly way to run AI models on your own machine: a desktop app that finds, downloads, and runs local models, then serves them to your editor and agents over a local API. It is free, even at work, and unlike a CLI-first tool it lets Claude Code connect straight to a local model with no proxy. Here is the whole thing, download to editor.

akira · 6 min

 A hand-inked terminal window with Aider pair-programming, a git branch beside it auto-committing each change, cables running out to model options labelled Claude, OpenRouter, and a local Ollama, the Okane Land starfish at the keyboard.

Tools & Stacks

How to use Aider: AI pair programming in your terminal

Aider is an open-source AI pair programmer that lives in your terminal, commits every change to git so you can always undo it, and points at almost any model, including a free local one. Here is the whole thing: install, run it in a repo, the commands you will use, and how to wire it to Claude, to OpenRouter, or to Ollama running on your own machine for zero per-token cost.

akira · 7 min

 A hand-inked laptop running a small local model, a llama sitting on the keyboard, an API port labelled 11434 wired out to an editor and a terminal agent, the Okane Land starfish plugging in the cable.

Tools & Stacks

How to use Ollama: run AI models on your own machine

Ollama runs open models on your own computer, for free, with one install command and one run command. For a vibe coder the appeal is blunt: inference that costs nothing, never leaves your laptop, and has no rate limit but your hardware. Here is the whole thing, install to editor, plus the plain catch about your RAM and the reason Claude Code will not point at it without help.

akira · 6 min

 A hand-inked control panel wired to a small site: a citation counter reading six, an index gauge climbing from five to thirty-six, and a server-log printout showing answer-engine bots skipping the llms.txt file, the mascot starfish reading the printout.

Marketing & Ads

How to make your product citable: an AEO walkthrough

The get-cited-by-ChatGPT genre is thick with unverifiable stats. We can only publish the walkthrough we can show our work for: our own agent-facing plumbing, tied to the files on disk, our Bing citation report going from zero to six, a fifteen-day server-log test of whether answer engines even read an llms.txt, and the tactics our own research made us drop. Every step carries either a before-and-after or a named source.

akira · 6 min

 A hand-inked context window drawn as a shelf, with CLAUDE.md sitting on it permanently, skills and subagents waiting in a drawer marked on demand, and hooks pinned outside the shelf entirely, the mascot starfish weighing each one.

Tools & Stacks

Hooks, skills, subagents, plugins, CLAUDE.md: what each is for

Five ways to extend Claude Code, and the confusion is real: hooks or skills, skills or subagents, CLAUDE.md or a skill. They differ on one axis that decides everything else, when each one enters the context window. Anthropic just published that axis and rebuilt around it. Here is the tested taxonomy, with the token cost of each measured on a clean session.

akira · 6 min

 A hand-inked two-panel: on the left a minimal MCP server as a small labelled box with one tool wired in, on the right six directory mailboxes with three flags up, one envelope stuck half-in, two question marks, the mascot starfish posting a letter.

Tools & Stacks

Ship an MCP server for your product

Most "how to build an MCP server" guides stop at hello-world and skip the part builders actually ask next: does anyone find it? We shipped one for this site, eleven days before the spec went final, so this walkthrough comes with real distribution receipts, including the submission that is still stuck. Here is the minimal build under the current spec, and the real reach.

akira · 6 min

 A hand-inked trail map from an idea lightbulb to a shipped app flag, passing six labelled gates: scope, tool, the 70% wall, deploy, security, users, the Okane Land starfish walking the path with a small pack.

Getting Started

How to build an app with AI from scratch

AI gets a non-coder from idea to a working demo quickly. Then the same person hits the same walls, in the same order: scope, tool choice, the 70% problem, the free-tier cliffs, security, and getting paid. This is the map of that whole path, with the real numbers at each wall and the deeper guide that gets you over it.

akira · 7 min

 A hand-inked split panel: on the left a closed box labelled forget the code exists, on the right the same box opened with three arrows reading it, explaining it, rebuilding it, the Okane Land starfish holding a pencil over the open box.

Skills & Craft

Can you actually learn while vibe coding?

By its coiner's own definition, pure vibe coding is the choice not to learn: forget the code exists. So learning starts the moment you break that definition on purpose. Here are the three cheap moves that turn a delegated diff into a mental model, and the reason your own sense that you understood it cannot be trusted.

akira · 7 min

 A hand-inked workflow canvas labeled one run equals one execution, with a coral trigger node branching through a step node and a coral AI node into a checked output, a dashed error lane to an alert bell, and an inked arrow to a client invoice card with a coral total, a monthly retainer chip, and a read-the-license footnote; the Okane Land seal and wordmark top-left and the starfish working from the lower left, the mascot of this n8n guide.

Tools & Stacks

How to use n8n, and when to write the code instead

n8n bills a workflow run as one execution, no matter how many steps it takes, and that pricing quirk is why the AI client-work trade settled on it. The earnings underneath are modest, recurring, and real, the guru numbers are mostly marketing, and the license has a clause most people selling builds have never read.

akira · 9 min

 A hand-inked prompt card with a coral spark turning, via an inked arrow, into an app window whose credit meter runs low, with a dotted exit arrow leading to a coral git branch labeled your repo; the Okane Land seal and wordmark top-left and the starfish watching from the lower left, the mascot of this Lovable guide.

Tools & Stacks

How to use Lovable, and when to take the code and go

Prompt in, working app out. The speed is real and front-loaded: the demo lands the same afternoon, the real costs land once you ship it as more than a demo, and the exit ramp, GitHub then Cursor, is part of driving it well.

akira · 9 min

 A hand-inked phone ad with a too-perfect AI spokesperson marked with an AI info chip, beside a hook matrix of six variant cards with one coral winner; the Okane Land seal and wordmark top-left and the starfish directing with a clapperboard from the lower left, the mascot of this AI UGC guide.

Marketing & Ads

How to make AI UGC ads, and when to pay a human instead

AI UGC tools sell a spokesperson video for the price of a coffee, and the working use case is narrower than the demos: high-volume hook testing. The money is in cheap volume, dozens of guesses for the price of one human clip, and the ceiling is trust: the moment an ad must carry a real claim or a real feeling, you hire the person.

akira · 9 min

 A hand-inked board of the vibe coding stack: two coral slot cards, the editor Cursor at twenty dollars a month and the agent Claude Code at twenty, beside dashed optional cards for app builders and same-slot twins, with an inked arrow to a coral price tag reading forty dollars a month before the meter moves.

Tools & Stacks

The vibe coding stack: what to pay for and what to skip

The whole vibe coding stack, slot by slot: what each one really costs, what the field runs by the numbers, the app-builder question, and the version that costs nothing.

akira · 6 min

 A hand-inked terminal window with a coral $ prompt and a command, three checkmarked steps, a plus-minus diff, and a coral block cursor; the Okane Land seal and wordmark top-left and the starfish working from the lower left, the mascot of this Claude Code guide.

Tools & Stacks

How to run Claude Code, and when to take back the keyboard

It ships on the same subscription as your Claude chats and drains the same usage pool. Delegation is the whole skill: given a bounded task and a real review it earns the price, unwatched it produces confident wreckage, and for quick in-editor edits Cursor stays the better buy.

akira · 9 min

 A hand-inked answer card with a coral spark in the ask bar, an answer carrying three coral numbered citations, and a row of three numbered source chips; the Okane Land seal and wordmark top-left and the starfish reading from the lower left, the mascot of this Perplexity guide.

Tools & Stacks

Where Perplexity earns its keep, and where it doesn't

Perplexity does the searching and the reading; the checking stays your job. A cited answer reads like a verified one and is not, and the tool pays for itself exactly as long as you keep clicking the footnotes.

akira · 8 min

 A hand-inked code-editor window with a few code lines and a coral text cursor, the Okane Land seal and wordmark top-left, and the starfish reviewing a diff, the mascot of this Cursor guide.

Tools & Stacks

How to actually use Cursor, and where it stops being worth it

The Tab autocomplete is the reason people stay; the credit meter is the reason they watch the dashboard. Tab by itself can carry the price, credits are the permanent catch, and the agent repays exactly the attention you give its diffs.

akira · 9 min