Missions

Why this matters: Before you can analyze subnets with Nexus, you should see what the network looks like visually. TaoMarketCap shows you the whole Bittensor ecosystem in your browser — no VPS needed.

What you're doing: Browsing the Bittensor network visually on your Mac, exploring subnets, and picking interesting ones to research later.

1. Open your Mac browser and go to taomarketcap.com
2. You'll see a list of all Bittensor subnets with prices and market data
3. Click on the "Subnets" page — browse through the list
4. Notice: some subnets have high prices (lots of people staking), others are cheap (less interest)
5. Click on any subnet that looks interesting — explore its detail page
6. Look for: number of miners, registration cost, price chart, description of what it does
7. Pick 3 subnets that caught your eye and write down their numbers — you'll analyze these in Nexus later

Done when: You browsed TaoMarketCap and picked 3 subnets that caught your eye.

Why this matters: The metagraph shows you every single miner in a subnet and how they're performing. This is the raw data behind Nexus's analysis — seeing it yourself helps you understand what Nexus is telling you.

What you're doing: Reading a metagraph table on TaoStats, identifying top and bottom performers, and comparing what you see to Nexus data.

1. Open taostats.io in your browser
2. Navigate to the metagraph page for any subnet (taostats.io/metagraph/[number] — try subnet 85 or one you picked in the previous mission)
3. You'll see a table of all miners with columns like: UID, incentive, emission, stake
4. Find the miner with the highest incentive score — this is the top performer
5. Find a miner with very low incentive — this one is struggling
6. Count roughly how many miners have decent scores vs how many have almost zero — this tells you the distribution (the Gini concept from the Bittensor page)
7. Compare what you see to what Nexus tells you about the same subnet — do they match?

Done when: You read a metagraph on TaoStats, found the top and bottom miners, and compared the data with Nexus.

Why this matters: Mining decisions aren't made from one snapshot — you need to watch trends. Is a subnet getting better or worse? Are new miners joining or leaving? A week of observation teaches you more than hours of reading.

What you're doing: Tracking 3 subnets for 5 days to observe trends in price, miner count, and Nexus alerts.

1. Pick 3 subnets to track (use ones you've already researched)
2. Every morning for 5 days, open taomarketcap.com and note for each subnet:
   • Alpha token price (up or down from yesterday?)
   • Number of miners (more or fewer?)
   • Any big changes visible on the chart?
3. Also check Nexus each morning: Are there any alerts for subnets [X], [Y], [Z]?
4. After 5 days, write a brief summary: which subnet improved? Which declined? Did any surprise you?
5. Share your findings with your team lead

Done when: You tracked 3 subnets for 5 days and wrote a summary of the trends you observed.

Why this matters: Warp is where you'll type the SSH command to connect to the VPS. It's like the built-in Terminal but much friendlier — auto-complete, better visuals, and less scary for someone who's never used a terminal before.

1. Open your browser and go to warp.dev
2. Click "Download for Mac"
3. Open the downloaded file and drag Warp to your Applications folder
4. Open Warp (Cmd+Space → type "Warp" → Enter)
5. Create an account when it asks (you can sign in with Google)
6. You should see a dark window with a blinking cursor — that's Warp ready to go

Done when: Warp is installed and you can open it from Applications.

Why this matters: Dark Reader makes every website dark mode — much easier on your eyes when you're staring at TaoMarketCap data tables for hours. LastPass saves all your passwords so you never forget them.

Dark Reader:

1. Open Chrome and go to the Chrome Web Store (search "Chrome Web Store" in Google)
2. Search for "Dark Reader"
3. Click "Add to Chrome" → "Add extension"
4. Click the puzzle piece icon in Chrome's toolbar → pin Dark Reader
5. Click the Dark Reader icon → toggle it ON — every website is now dark mode

LastPass:

1. Go to lastpass.com and click "Get LastPass Free"
2. Create an account with your email
3. Choose a strong master password — this is the ONE password you must remember
4. In Chrome Web Store, search "LastPass" → Add to Chrome
5. Log in to the extension — from now on, LastPass saves and fills passwords automatically

Done when: Dark Reader is on (websites look dark) and LastPass is installed and logged in.

Why this matters: Rectangle lets you snap windows side by side — essential when you want TaoMarketCap on the left and Discord on the right. Bookmarks give you one-click access to the sites you'll use every day.

Rectangle:

1. Go to rectangleapp.com and click "Download"
2. Open the file, drag Rectangle to Applications
3. Open Rectangle — it asks for accessibility permissions, click "Open System Preferences" and enable it
4. Test it: open any window and press Ctrl+Option+← to snap it to the left half of your screen
5. Open another window and press Ctrl+Option+→ — now you have two windows side by side

Bookmarks:

1. In Chrome, go to taomarketcap.com and press Cmd+D to bookmark it
2. Create a folder called "Bittensor" and save the bookmark there
3. Do the same for taostats.io and your training site (training.nurax.io)
4. Now you can access all three with one click from Chrome's bookmark bar

Done when: Rectangle works (you can snap windows with keyboard shortcuts) and your 3 bookmarks are saved.

Why this matters: Discord is where you catch information before it appears anywhere else. Developers post updates, miners report issues, and opportunities surface here first.

What you're doing: Setting up Discord on your Mac, joining the main Bittensor community, and connecting with your team lead.

1. Download Discord from discord.com/download and install on your Mac (or use it in your browser)
2. Create an account if you don't have one
3. Join the main Bittensor Discord: discord.gg/bittensor
4. Browse the channel list on the left — find #general, #mining, and any subnet-specific channels
5. Add your team lead as a friend: click Friends → Add Friend → type nicknames96 → Send request

Done when: You're in the Bittensor Discord and your team lead is in your friends list.

Why this matters: Each subnet has its own community. Joining a subnet's Discord gives you insider access to dev updates, scoring changes, and community sentiment — data Nexus can't always capture.

What you're doing: Finding and joining a subnet Discord server, then reading their announcements to get a feel for the team's activity.

1. Pick a subnet you've been researching (from TaoMarketCap or Nexus)
2. Find its Discord: check the subnet's GitHub repo README, or ask Claude Code "Find the Discord server for Bittensor subnet [number]"
3. Join the server
4. Find the #announcements channel and read the last 5-10 messages
5. Note: who are the developers? Are they active? When was the last update?

Done when: You've joined at least one subnet Discord and read their announcements.

Why this matters: The VPS is the remote server where all your tools live -- Agent Deck, Claude Code, Nexus, everything. Connecting to it is like unlocking the front door. Until you can do this, you can't access anything. As an AI Architect, you'll connect to this server every time you start work.

What you're doing: Using SSH (a secure remote connection) from your Mac's Terminal app to connect to the VPS. Once connected, you'll launch Agent Deck.

1. Open Terminal on your Mac (press Cmd+Space, type "Terminal", press Enter). You'll see a dark window with a blinking cursor -- that's your local terminal.
2. Type the SSH command from the Access & Security page and press Enter. It looks like ssh giorgia@207.180.228.169. You're telling your Mac to connect to the remote server.
3. Enter the password when prompted. You won't see any characters as you type -- that's normal security behavior, not a bug. Just type the password and press Enter.
4. You should see the server prompt change -- it will show the server name instead of your Mac name. This means you're connected. Type agent-deck and press Enter.
5. You're in. You should see your groups (Work, Nexus) on the left side of the screen.

Done when: You see the Agent Deck home screen with groups and sessions listed.

Why this matters: Agent Deck is your control center. Every tool you'll use as an AI Architect launches from here. Learning to move around it quickly is like learning where the apps are on your phone -- once it's automatic, you stop thinking about it and focus on the actual work.

What you're doing: Practicing switching between groups and sessions until it feels natural. You'll open Claude Code, go back, open Nexus, go back -- over and over.

1. Use the arrow keys (up/down) to highlight the Work group, then press Enter. You'll see the sessions inside that group appear on the right side.
2. Select the personal session and press Enter. Claude Code opens -- you'll see a text interface where you can type questions and instructions.
3. Press Ctrl+Q to go back to Agent Deck. Claude Code keeps running in the background -- you haven't lost anything.
4. Now navigate to the Nexus group, select chat, and press Enter. Nexus Intelligence loads -- it looks similar to Claude Code but it's connected to Bittensor subnet data.
5. Press Ctrl+Q to go back to Agent Deck again.
6. Do this 3 more times (Work → Claude Code → back, Nexus → chat → back) until it feels automatic. Speed matters less than confidence.

Done when: You can switch between Work and Nexus groups without hesitation, and Ctrl+Q feels natural.

Why this matters: Every new project you work on gets its own session in Agent Deck. When you start analyzing a new Bittensor subnet or building a new tool, you'll create a session for it. This keeps your work organized -- just like creating a new folder on your Mac for each project.

What you're doing: Creating a brand-new group and session from scratch, then using Claude Code inside that session to create your first file.

1. Open Agent Deck. Press G to create a new group. A popup appears -- type Practice and press Enter. You'll see the new "Practice" group appear in the list on the left.
2. Inside the Practice group, press N to create a new session. A form appears with fields to fill in.
3. Name the session test-project -- this is just a label so you know what the session is for.
4. Set the path to /home/giorgia/test-project -- this is the folder on the server where your project files will live. If the folder doesn't exist yet, Agent Deck creates it for you.
5. Select claude as the tool -- this means the session will open Claude Code when you enter it.
6. Press Enter to confirm. The new session appears in the Practice group. Open it by selecting it and pressing Enter.
7. In Claude Code, type: Create a file called hello.txt with the text 'My first file' and press Enter. Claude creates the file on the server for you.
8. Type: Show me the contents of hello.txt -- Claude reads the file back to you, confirming it was created with the text you specified.

Done when: You created a new group called "Practice", a new session called "test-project", and Claude created a file for you inside it.

Why this matters: Claude Code is the AI that builds things for you -- it reads files, writes code, creates projects, and manages the server. This is your first real interaction with it. As an AI Architect, most of your work happens through conversations with Claude Code, so getting comfortable talking to it is essential.

What you're doing: Asking Claude Code two simple questions to see how it responds. You'll ask where you are on the server and what files exist there.

1. Open Agent Deck, navigate to Work → personal and press Enter. Claude Code opens.
2. Type: What directory am I in right now? and press Enter. Claude will respond with a path like /home/giorgia -- that's the folder you're currently working in on the server.
3. Read the answer. Claude tells you the current folder path. Think of directories as folders on your Mac -- same concept, just on the server.
4. Now type: List all the files in this directory and press Enter. Claude will show you a list of files and folders.
5. Read the file list. You just used Claude Code to explore the server. Each item is either a file or a folder that exists on the VPS.

Done when: Claude responded to both questions and you understood the answers -- you know what directory you're in and what files are there.

Why this matters: Before you can analyze subnets, you need to understand the network. Claude Code is your teacher -- you can ask it anything and it will explain in plain language, at whatever level you need.

What you're doing: Having a conversation with Claude Code to build your understanding of what Bittensor is, how mining works, and what TAO (the network's currency) is all about.

1. Open Agent Deck → Work → personal and press Enter. Claude Code opens.
2. Ask Claude: Explain Bittensor to me in simple terms. What is it, how does mining work, and what is TAO?
3. Read the response carefully. Then ask follow-up questions about anything you didn't understand -- there are no bad questions here.
4. Ask: What is a subnet and why do different subnets exist?
5. Keep asking until the big picture makes sense to you. You should be able to explain in your own words: what Bittensor is, what miners do, and why subnets matter.

Done when: You can explain in your own words what Bittensor is, what TAO is, and why subnets exist.

Why this matters: Real data teaches better than theory. You'll see actual subnets, real emissions, real miners -- the live state of the Bittensor network right now.

What you're doing: Using Nexus Intelligence to query live subnet data, see which subnets pay the most per miner, and read a full subnet analysis report.

1. Open Agent Deck → Nexus → chat and press Enter. Nexus Intelligence loads.
2. Ask Nexus: Show me the top 10 subnets by emission per miner
3. Look at the results -- which subnet pays the most per miner? How many miners are competing in each one? Notice how different subnets have very different numbers.
4. Ask Nexus: What does subnet 85 do? Give me the full report
5. Read the verdict: is it MINE, SKIP, or WATCH? What are the reasons? Look at the emission, miner count, and any warnings Nexus mentions.

Done when: You explored the top subnets by emission and read a full report for subnet 85. You understand what the verdict means.

Why this matters: The quality of what Claude builds depends entirely on how you ask. A vague prompt gets a generic result. A precise prompt gets exactly what you want. This is the single most important skill for an AI Architect -- you're not writing code, you're writing instructions that make AI write the right code.

What you're doing: Comparing a vague prompt with a precise one so you can see the difference with your own eyes. You'll use the 5-part prompt structure from the Prompt Building guide.

1. Open Claude Code (Work → personal in Agent Deck).
2. First, try a vague prompt: type Make a webpage and press Enter. Claude will generate something -- but it's generic and probably not what you had in mind.
3. Read what Claude generates. It works, but it's not specific to your needs. This is what happens when you don't give clear instructions.
4. Now try a precise prompt using the 5-part structure: Create an HTML file called status.html. It should show a simple table with 3 rows: Subnet 18 (status: mining), Subnet 42 (status: watching), Subnet 7 (status: stopped). Use a dark background (#0f0f13) and white text. Add a title 'Mining Status' at the top.
5. Compare the two results. The precise prompt gives you exactly what you wanted -- specific subnets, specific colors, specific layout. This is the power of structured prompting.

Done when: You can see the difference between a vague prompt result and a precise one, and you understand why specificity matters.

Why this matters: Plan Mode lets you review what Claude is about to do before it does it. This is critical when you're working on real projects -- you don't want Claude making changes you haven't approved. As an AI Architect, reviewing plans before execution is part of your job. You're the decision-maker, Claude is the builder.

What you're doing: Turning on Plan Mode, giving Claude an instruction, reviewing the plan it proposes, and then approving it. You'll see exactly what changes Claude intends to make before any file is modified.

1. Open Claude Code (Work → personal in Agent Deck).
2. Type: /plan and press Enter. This toggles Plan Mode on. You'll see a confirmation message that plan mode is now active.
3. Now give an instruction: I want to add a footer to status.html that shows the current date and a copyright notice
4. Instead of immediately making changes, Claude shows you a plan -- what files it will change, what it will add, and what the result will look like. Read through it carefully.
5. If the plan looks good, type yes or go ahead and press Enter. Claude executes the plan and shows you the result.
6. If the plan doesn't look right, tell Claude what to change: Actually, don't include the date, just the copyright notice. Claude will revise the plan.
7. Type: /plan again to turn Plan Mode off for normal use. Now Claude will go back to making changes immediately.

Done when: You used Plan Mode to review changes before they happened, and you understand the difference between plan mode and normal mode.

Why this matters: The Gini coefficient tells you if a subnet is fair or if one big miner takes everything. This is critical for deciding where to mine -- a high Gini means a newcomer will earn almost nothing because the top miners dominate the rewards.

What you're doing: Comparing fair subnets (low Gini) with winner-take-all subnets (high Gini) to see the difference, and understanding why this matters for someone who rents GPU hardware.

1. In Nexus (Nexus → chat), ask: Show me subnets with Gini below 0.3
2. These are the FAIR subnets where rewards are shared relatively equally among miners. Notice how many miners are in each one.
3. Now ask: Show me subnets with Gini above 0.8
4. These are winner-take-all subnets -- dangerous for newcomers because a few top miners capture almost all the rewards.
5. Ask: Why does Gini matter for a cloud miner who rents GPUs?
6. Read the explanation. When you rent GPUs, you're paying real money every hour -- if a subnet's Gini is too high, you'll spend more on rent than you earn in rewards.

Done when: You compared low-Gini and high-Gini subnets and can explain why a high Gini is risky for a new miner.

Why this matters: Mining decisions are always comparisons -- is subnet A better than subnet B for your hardware and budget? Learning to compare systematically instead of guessing is what separates a good AI Architect from someone who just picks randomly.

What you're doing: Asking Nexus to compare two specific subnets side by side, evaluating multiple metrics, and making a reasoned decision about which one you'd mine.

1. In Nexus, ask: Compare subnet 85 and subnet 111. Which is better for a miner who rents GPUs?
2. Look at the comparison: emission per miner, Gini, miner count, hardware requirements. Which subnet looks more attractive at first glance?
3. Ask: What hardware does each subnet need? Can I run either one on a rented GPU?
4. Read the hardware requirements. Some subnets need specialized GPUs or large amounts of VRAM -- this affects your rental costs and whether it's even possible.
5. Make your decision: which subnet would you mine? Write down 3 reasons why. Your reasons should reference specific numbers from the comparison (like emission, Gini, or hardware cost).
6. Show your reasoning to your team lead

Done when: You compared two subnets, made a decision, and can explain your reasoning with 3 specific data points.

Why this matters: Nexus Intelligence is the research tool that tells you which Bittensor subnets are worth mining. As an AI Architect, one of your main responsibilities is analyzing Nexus reports and making informed recommendations. Learning to read these reports -- understanding the verdict, the numbers, and what they mean -- is core to your role.

What you're doing: Opening Nexus and asking it to analyze a subnet. You'll learn to read the key metrics: the verdict (MINE/SKIP/WATCH), emission per miner, miner count, and the Gini coefficient (how fairly rewards are distributed).

1. Open Agent Deck, switch to Nexus → chat and press Enter. Nexus Intelligence loads.
2. Ask: Give me a quick report on the most active Bittensor subnet right now and press Enter. Nexus will analyze its database and respond with a structured report.
3. Read the report. Look for these key elements: the verdict (MINE means worth mining, SKIP means avoid, WATCH means keep an eye on it), emission per miner (how much TAO each miner earns), and miner count (how many competitors there are).
4. Ask a follow-up: What's the Gini coefficient for this subnet? Is it fair?
5. Read the answer. High Gini (close to 1.0) means rewards are concentrated at the top -- a few miners get most of the TAO. Low Gini (close to 0) means rewards are spread more evenly across miners. For new miners, low Gini subnets are generally better.

Done when: You read a Nexus report and understood the verdict, emission per miner, miner count, and Gini coefficient.

Why this matters: Real mining decisions involve predicting the future -- what happens if more miners join? What if emissions change? Nexus can model these scenarios so you make decisions based on projections, not guesses. As an AI Architect, your recommendations to your team lead need to be backed by data and analysis, not just gut feeling.

What you're doing: Asking Nexus to model two different scenarios for the same subnet -- the current state and a hypothetical future -- then comparing them to understand how changing conditions affect profitability.

1. In Nexus (Agent Deck → Nexus → chat), ask: If I mine subnet 18 with one RTX 4090, what's my expected daily emission? Is it profitable?
2. Read the analysis. Note the projected emission (how much TAO per day), the break-even estimate (how long until you earn back the hardware cost), and any risks mentioned (like increasing competition or changing incentive mechanisms).
3. Now ask: What if 50 more miners join subnet 18 in the next month? How does that change my profitability?
4. Compare the two scenarios. More miners means the same total rewards split among more participants -- your share shrinks. This is how you make decisions before spending money on hardware or registration fees.

Done when: You ran a what-if scenario and compared two different outcomes, and you understand how increasing competition affects profitability.

Why this matters: Before spending money on GPU rentals, you simulate different scenarios to find the best strategy. This is like planning a budget -- you test different options on paper before committing real money.

What you're doing: Asking Nexus to model different budget scenarios and recommend mining strategies for each one. You'll see how budget constraints change which subnets are viable.

1. In Nexus, ask: What if I had a budget of $500 per month for GPU rentals? Which subnets should I mine?
2. Read the recommendations -- Nexus will consider hardware costs versus emission income to suggest the best subnets for that budget.
3. Now ask: What if I only had $200 per month?
4. Compare the two scenarios -- how does a smaller budget affect which subnets are viable? Some subnets might drop off the list entirely because their hardware costs are too high.
5. Ask: Are there any CPU-only subnets I could mine without renting GPUs?
6. Read the answer. CPU-only subnets have zero GPU rental cost, which changes the economics completely -- but they usually have lower emissions too.

Done when: You compared at least two budget scenarios and can explain how budget affects which subnets are worth mining.

Why this matters: GSD (Get Shit Done) is the project management workflow that turns vague ideas into structured plans and working software. Instead of giving Claude one instruction at a time and hoping for the best, GSD breaks your project into phases and plans that Claude executes systematically. As an AI Architect, you'll use GSD for every real project.

What you're doing: Initializing a real project using GSD commands. You'll describe what you want to build, and GSD will create a structured roadmap with phases, plans, and tasks. Then you'll walk through the first phase discussion and planning steps.

1. Open Claude Code (Work → personal in Agent Deck).
2. Type: /gsd:new-project and press Enter. This starts the project initialization wizard.
3. When Claude asks what you're building, describe it clearly: A simple monitoring page that shows the status of our Bittensor mining operations. It should be a static HTML page with a dark theme that reads data from a JSON file.
4. Answer Claude's follow-up questions -- it will ask about technology (say HTML, CSS, and JavaScript), features (say a table showing subnet name, status, and emission), and constraints (say no build step, static files only).
5. Claude creates a .planning/ directory inside your project folder. This directory contains your project roadmap, requirements, and state tracking. You'll see Claude list the files it created.
6. Type: /gsd:discuss-phase to start a discussion about the first phase. Claude will ask questions about what you want in phase 1 -- answer naturally, like you're explaining the project to a colleague.
7. After the discussion, type: /gsd:plan-phase to create the formal plan. Claude generates a detailed plan with specific tasks, file lists, and success criteria.

Done when: You have a .planning/ directory with a roadmap, and at least one phase plan created.

Why this matters: This is the complete workflow you'll use in your actual job. It combines every tool you've learned -- Nexus for research, Claude Code for technical analysis, and GSD for structured execution. This is what an AI Architect does: research opportunities with Nexus, dive deep into the code with Claude Code, and deliver structured analysis using GSD. If you can do this mission, you can do the job.

What you're doing: You'll research subnets with Nexus, pick one to investigate, clone its code repository with Claude Code, and use GSD to systematically analyze how it scores miners. Each step builds on the previous one.

Step 1 -- Research with Nexus

1. Open Agent Deck and navigate to Nexus → chat. Nexus Intelligence loads.
2. Ask: Which Bittensor subnets are currently worth investigating for mining with an RTX 4090? Give me your top 3 with brief reasons.
3. Nexus responds with a list of subnets. Each entry includes the subnet number, what it does, why it's interesting, and key metrics. What you're looking for: subnets with a MINE or WATCH verdict, reasonable emission per miner, and a Gini coefficient that isn't too high (meaning rewards aren't monopolized by a few big miners).
4. Pick one subnet from the list. Write down its number (like "subnet 18") and its repository URL (a GitHub link that Nexus provides). You'll need both for the next steps.

Step 2 -- Clone the repository

1. Press Ctrl+Q to go back to Agent Deck. Navigate to Work → personal and press Enter. Claude Code opens.
2. Tell Claude: Clone the repository for Bittensor subnet [NUMBER] from [URL] -- replace [NUMBER] and [URL] with the actual subnet number and GitHub link you noted. "Cloning" means downloading the entire codebase from GitHub to your server so Claude can read and analyze it.
3. Claude will download the repository. You'll see it running a git clone command and reporting the files it downloaded. When it's done, the subnet's entire codebase is on your server in a new folder.

Step 3 -- Set up GSD analysis

1. Tell Claude: /gsd:new-project and describe it as: Analyze subnet [NUMBER] to understand its scoring mechanism and mining requirements
2. GSD creates a .planning/ directory inside the cloned repository. This sets up structured analysis instead of random exploration.
3. Run /gsd:discuss-phase and tell Claude you want to map the codebase and understand the scoring system. This is the most important part -- the scoring system determines how miners earn rewards, and understanding it is the key to deciding whether to mine.

Step 4 -- Execute the analysis

1. Run /gsd:plan-phase -- Claude creates a structured plan for analyzing the codebase.
2. Run /gsd:execute-phase -- Claude systematically goes through the code, mapping the architecture, identifying the scoring mechanism, and documenting what it finds.
3. Read the analysis. Claude will produce a summary explaining: how the subnet scores miners, what hardware/software is needed, what the scoring criteria are, and whether there are any red flags (like penalties for poor performance or high barrier to entry).
4. You now have a data-driven analysis of whether this subnet is worth mining. This is the deliverable you'd present to your team lead.

Done when: You used Nexus to find a subnet, cloned it with Claude Code, and used GSD to analyze the scoring system. You can explain what the subnet does and how it scores miners.

Why this matters: Building something and making it accessible from your browser proves you can use Claude Code for real deployments -- not just exercises. As an AI Architect, you'll create dashboards, monitoring pages, and internal tools. This mission shows you the full cycle: describe what you want, Claude builds it, and you can see it live in your browser from your Mac.

What you're doing: Asking Claude Code to create a web page, set up the server to serve it, and then opening it in your Mac's browser to verify it works. You'll go from "nothing" to "live website" in a few minutes.

1. Open Claude Code (Work → personal in Agent Deck).
2. Ask Claude: Create a simple HTML page at /var/www/html/giorgia-test/index.html that shows my name, today's date, and a message 'I built this with Claude Code'. Use a dark background and purple accents.
3. Claude creates the file. You'll see it write the HTML code and confirm the file was created at the specified path.
4. Now ask: Set up nginx to serve this page on port 9999. Create the nginx config, enable it, and restart nginx. Nginx is the web server that makes your page accessible from the internet -- Claude handles all the configuration for you.
5. Open your Mac browser (Safari or Chrome) and go to: http://207.180.228.169:9999
6. You should see your page -- dark background, purple accents, your name, and the message. You just deployed a website using only Claude Code. No coding knowledge required.

Done when: You can see your custom page in your Mac's browser at port 9999.

Why this matters: This is the final test. No instructions, no hand-holding. You choose what to build and you build it using everything you've learned. This proves to yourself (and to your team lead) that you can work independently as an AI Architect. If you can complete this mission, you're ready for real projects.

What you're doing: Choosing your own project idea, planning it with GSD, building it with Claude Code, deploying it on the VPS, and showing the result. The entire process from idea to live website, on your own.

1. Think of something you want to build. It doesn't need to be complex -- the goal is to practice the full workflow independently. Ideas: a personal bookmark page with your favorite links, a subnet comparison table showing the top 5 subnets, a daily work log page, a recipe collection, or a simple dashboard showing mining stats.
2. Open Claude Code (Work → personal in Agent Deck). Create a new session in Agent Deck for this project (press Ctrl+Q, then N to create a new session in the Work group).
3. Use GSD to plan and build it: type /gsd:new-project and describe what you want to build in detail. Remember the 5-part prompt structure -- be specific about content, appearance, and structure.
4. Walk through all four GSD phases: /gsd:discuss-phase (talk about what you want), /gsd:plan-phase (create the plan), /gsd:execute-phase (build it), and check the results.
5. When it's done, deploy it on the VPS so you can access it from your browser. Ask Claude to set up nginx for your project, just like you did in the previous mission.
6. Open it in your Mac's browser to verify it works. Then show your team lead what you built -- he'll be impressed.

Done when: You built and deployed something entirely on your own, from idea to live website, and you can show it to your team lead.

Why this matters: Understanding how miners are scored is the single most valuable analysis you can do. It tells your team lead exactly what a miner needs to do to earn the most TAO.

1. Open Agent Deck → create a new session in the Work group (press N)
2. Name it after any subnet you want to analyze (pick one with a MINE or WATCH verdict from Nexus)
3. In Claude Code, ask it to clone the subnet's GitHub repo
4. Ask Claude: Analyze the scoring function in this repository. How are miners evaluated? What determines who gets the highest rewards?
5. Ask Claude: What type of competition is this -- model quality, data quality, compute proof, or hybrid?
6. Ask Claude: What would a miner need to do to get the best possible score?
7. Write down the key findings -- you'll use these in the next mission

Done when: You analyzed a subnet's scoring function and can explain how miners are evaluated and what determines the highest rewards.

Why this matters: Before spending money on GPU rentals, you need to calculate if a subnet will actually make money. This is the difference between smart mining and gambling.

1. Use the subnet you analyzed in the previous mission
2. In Nexus, get: emission per miner, current TAO price, miner count
3. In Claude Code, ask: Based on the hardware requirements for this subnet, how much would it cost per month to rent a GPU from Vast.ai or Targon?
4. Calculate: Monthly TAO earned = emission per miner × 7200 (tempos per day) × 30 days
5. Calculate: Monthly cost = GPU hourly rate × 24 × 30
6. Calculate: Monthly profit = (TAO earned × TAO price) - Monthly cost
7. Is it profitable? By how much? What's the risk if TAO price drops 20%?

Done when: You calculated the profitability of mining a subnet including hardware costs, TAO earnings, and risk if TAO price drops.

Why this matters: This is your actual job output -- a complete analysis that Your team lead uses to decide where to mine. This mission combines everything you've learned.

1. Pick a subnet (use one you've already analyzed, or start fresh)
2. Gather ALL data: Nexus report + code analysis + profitability calculation
3. Write a recommendation using this structure:
   1. Subnet number and what it does (1-2 sentences)
   2. Scoring system summary (how miners are judged)
   3. Hardware requirements and estimated monthly cost
   4. Estimated monthly TAO earnings
   5. Risk assessment (Gini, competition, community health)
   6. Your verdict: MINE / SKIP / WATCH with 3 supporting reasons
   7. Any concerns or unknowns
4. Ask Claude Code to help you format this nicely
5. Send it to your team lead for review and feedback

Done when: You wrote a complete mining recommendation with all 7 sections and sent it to your team lead for review.