Unlock AI guides by Nate B. Jones
Browse the summaries, then sign in to read a member guide.
Agents
10 guidesBusiness in a Bot
Turn an idea into a small business that a few Bots run on one shared computer. You approve every important step. Start with the Super Doer guide.
Super Doer Bot
The first Bot I would start with. Stop narrating work and start preparing it. Name a teammate, hand it a real job, and open the file it left behind.
Build My AI Team
Download the Agent Builder, give it one goal with repeated work, and run a first real task. Keep every agent's instructions, sources, state, and work in folders you control.
The Agent Maintenance Loop
A step-by-step guide to check an AI agent's job, information, memory, tools, access, proof, and value before you keep, change, pause, or retire it.
Find a real job for your first AI agent
Document a repeated problem, check the pattern against real cases, decide whether an AI agent fits, and test a draft that a person reviews before anything is sent or changed.
Clean My AI Harness: Mission Fit
Review recent AI agent work, find repeated problems, and test each approved fix before changing the agent's setup.
The One-Minute Test
Route any task to chat, one accountable agent, a small agent team, or the manual path in about a minute, with a next action for whichever route wins.
Automation Discovery: Find work worth automating
Install Automation Discovery. It reviews records of your past work, checks each finding against the original record, and suggests two to five automations for you to choose from.
Astra: What work can I give this?
Learn to turn real work into a clear assignment. Six moves, a manager loop, and 23 worked recipes that teach you to write and steer your own prompts.
Fable or Astra: How to choose between frontier models
Which model I'd choose for building an app, doing research, making a spreadsheet, or thinking through a hard decision, and when using both is worth it.
Codex
4 guidesThe Ultimate Guide to Codex
Install OpenAI Codex, choose safe settings, set up a real project, inspect the work, attach notes to a page, delegate tasks, and save reusable instructions.
Use the Codex side panel
Use Files, Side chat, Review, Terminal, and Browser to inspect Codex's work before you accept it.
Codex browser and annotations
Attach a note to part of a page, send one change or several related changes, and verify that Codex edits the right file.
Use Codex threads and child threads
Give each Codex thread one outcome, choose whether a new message changes the current work or waits, delegate separate steps, define a long-running goal, and move unfinished work into a fresh thread.
General
27 guidesOwn your AI memory in about an hour
Move your memory, files, and instructions out of any single AI product: set up Open Brain, write one portable instruction file, migrate what your tools already know, and pass the switch test.
The $20 / $60 / $200 decision
Decide what to pay for AI at every budget: one main provider at $20, the three-tool split at $60, and the payback bar every plan must clear past $200.
Visible delegation: let a cheap model do the typing
Give a lower-cost coding agent one clearly defined task in a visible session while your strongest model plans the work, reviews every changed line, and reruns every check.
Run GLM-5.3 in Claude Code and Codex
Use Z.AI's $18 GLM Coding Plan for separate Claude Code and Codex sessions, then compare safer delegation, an OpenRouter fallback, and cost per accepted result.
Build the App: the setup guide
Choose a software type and one of two build routes. Set up the accounts, data storage, project files, publishing, security, and one working path.
Working Context Starter Kit
Build a small, current working context set for any AI project with four plain-text files, three copyable prompts, and the bare/focused/full test.
Make any skill your own
Learn how to read, test, edit, and move an AI skill between supported tools by working through a plain-text meeting-recap example.
Open Stack
Choose the Open Stack part that fixes your current problem. Start with Open Skills, Open Brain, or Open Engine, then add more only when you need it.
Open Brain: one memory for every AI
Learn how Open Brain stores shared memory, connects supported AI tools, and guides you through setup without writing code.
Ringer: Multi-agent swarms without the frontier bill
Run multi-agent swarms on inexpensive models. Your best model plans and reviews, cheap workers build in parallel, and automated checks verify every result.
How to Run a Chinese-Model Bakeoff in Ringer
Give a Chinese model and your current model the same real job, the same sources, and the same pass/fail check, then see which one earns the work.
Clean My AI Harness
Install the Claude or Codex edition, scan your AI setup without changing existing files, review one report, approve changes by number, and use the rollback path if something goes wrong.
Build your own token-burn dashboard
Track where your tokens go across Codex, Claude, and ChatGPT, then use the burn rate as a fluency signal.
Build a Healthcare Claim Appeals Agent
Assemble Open Skills primitives into an agent that turns your denial letter, EOB, and plan documents into a cited appeal packet.
Build a Tax Prep Organizer Agent
Use the same document-grounded primitives to turn your tax documents and real IRS sources into a structured, CPA-ready prep packet.
Build an Email Follow-Up Agent
Point the same document-grounded primitives at your inbox: a live MCP ingest loop that reconstructs threads, keeps a commitments ledger, and drafts cited follow-ups that stop at the send boundary.
Build a steering profile for any model
Record a model-specific rule, compare results with and without it, and keep only rules the evidence supports.
Audit an AI-visibility dashboard before you pay
Ask where each dashboard number came from, what the company counted, and what evidence it can show before you pay.
Clean sensitive documents without uploading them
Download Gemma 4 E2B with LM Studio and create a proposed redacted copy, which replaces sensitive details with labels, without sending the original to an online artificial intelligence service.
AI Airlock: clean sensitive documents before using AI
Install AI Airlock on a Mac with an M1 chip or newer, review every proposed removal, create a checked copy, and use that copy with an AI tool.
15 ways to cut token waste in Codex and Claude
Reduce the text Codex and Claude process for each job with 15 moves that cut reused input, large source files, long answers, and model calls you do not need.
The computer that fits you: what to change on your Mac or PC, and how to try Omarchy
What you can ask an agent to change on the Mac or Windows machine you already own, how far to let it reach, and the three ways to try Omarchy itself.
Claude Fable 5.1: exact prompts and longer workflows
Use the prompts and longer workflows behind Nate's Fable 5.1 video for spreadsheets, presentations, Word documents, prose, Blender, and price.
Hand your subscriptions to an assistant, and check its work
Nate asked Muse to go through his subscriptions and it found $5,350 a year in recurring spending. Here is the same job as prompts you can paste into Muse, Claude, or ChatGPT: build the list, decide what stays, cancel cleanly, and confirm it stopped.
Connect Nate's Library MCP to your AI app
Connect Nate's Library, the MCP server for Nate's posts, prompt kits, and guides, to Claude, Claude Code, Codex, Perplexity, Gemini, or Grok in about two minutes.
Find the Jev-shaped problems in your software
Get Jev running in about five minutes, find the Jev-shaped work in what you already have with one prompt, then hand your coding agent one of four recipe cards to build it.
The real assignment test: find out if a new model is better at your work
Nate's Lego build used 1% of his weekly Claude usage, but that number describes his job, not yours. Three prompts build your own repeatable test from the work you already do, so every new model gets graded on your assignments.