The Underground
Learn it by
building it.
Short, practical missions for the most useful free tools on GitHub. A self-paced underground: start any day, go at your own speed.
Start learning freeTRACK 00
Level 0: The basics
The two skills behind every tool on this site: installing from GitHub safely, and finding your way around the terminal.
- 5 missions
- 2h 2m
- Beginner
- 01
Install any free tool from GitHub, safelyThe one skill behind every tool on this site. Twenty minutes, no coding, and by the end you'll have downloaded a real project and run it yourself. - 02
Stop being scared of the black window: the 10 terminal commands every guide assumesEvery GitHub README skips the basics. Learn ten commands and three keyboard tricks in a practice folder, then finish by pulling a live weather forecast straight into your terminal. - 03
Give every project an undo button: Git without memorising a single commandYour AI coder just broke the thing that worked an hour ago, and with Git that's one key to undo. Git is the save-and-undo system behind every project on GitHub, and lazygit puts it all on one screen with single-key shortcuts, so there's nothing to memorise. Practise in a throwaway folder, then use it on anything. - 04
Make your terminal look like a pro's (and actually more useful)Oh My Posh swaps PowerShell's plain prompt for a colourful status bar: your folder, your Git branch and changes, how long the last command took and whether it failed. fastfetch adds a system-info banner on top. Free, no account, and every change can be undone. - 05
Install apps on a locked-down laptop, no admin password neededThe laptop says "ask your administrator" every time you try to install something. Scoop installs thousands of free apps into your own user folder instead: no setup wizards, no admin pop-ups, and one command updates the lot.
TRACK 01
Own your AI
Your own private ChatGPT on your laptop. Offline, free, nothing leaves your machine.
- 31 missions
- 34h 56m
- Beginner to Advanced
- 01
Run a ChatGPT-style AI on your own laptopGoogle's Gemma 4, running on your machine with Ollama. No account, no subscription, and it keeps answering with the Wi-Fi switched off, because nothing you type leaves your computer. - 02
The tiny AI model that got basic maths wrong, and the one to use insteadAsk a tiny free model when a 14:35 train arrives after 2 hours 50 and it confidently says 15:25. Find the model that gets it right on your laptop, plus a one-command test that checks any model before you trust it. - 03
Your own private ChatGPT, in your browserPut Open WebUI on top of Ollama and you get the ChatGPT experience: chat history, file uploads, voice, and model switching. All of it running on your own machine. - 04
Find the catch in your contract before you signDrop a tenancy agreement, payslip or bank letter into your own AI and get the dates, costs and catches back in plain English. The file never leaves your laptop. - 05
Translate letters and documents without uploading themTranslate your landlord's letter without sending it to Google. Your own translator on your laptop: messages, letters and whole Word or PDF files, offline, with nothing sent to a server. - 06
Make AI images on an ordinary laptop, no graphics cardType a sentence, get a picture in about a second. FastSD CPU runs image models on the laptop you already own: no NVIDIA card, no account, no credits, and your prompts never leave your machine. - 07
Put AI in your terminal: summarise files, logs and web pages with one commandCopy any error message, type one short command, and get the cause and the fix in plain English. LLM sends any text to the AI on your own laptop, so your notes, errors and logs never leave your machine. - 08
Turn any Word, Excel, PowerPoint or PDF into text your AI can actually readMicrosoft's free MarkItDown turns office files into clean Markdown, with the headings, tables and speaker notes kept. Paste it into ChatGPT, Claude or your own local AI. It runs on your laptop and uploads nothing. - 09
A 24/7 AI assistant that actually does things, running on your own PCOpenClaw is the open-source assistant from all those reels: it reads and writes files, runs commands and works on a schedule. Set it up on your own PC, lock it down so it can't wreck anything, and give it real jobs. Free with a local AI, or fast with a cheap API model. - 10
Feed any website to your AI in one commandOne command turns a web page into clean Markdown, with the menus and cookie banners stripped out. Save whole docs sections, pull tables off pages as data, and ask your own local AI about what you saved, all with free tools on your laptop. - 11
Pull the tables out of any PDF into Excel, with free AI that runs on your laptopDocling, the free document reader started at IBM Research, finds the tables in a PDF (even a scanned one) and rebuilds them as a spreadsheet. It runs on your own machine, so bank statements and invoices never get uploaded anywhere. - 12
Cancel Copilot: a free, private coding assistant in VS Code on your own laptopNo subscription, no usage limits, no code sent to anyone. Continue puts chat, one-instruction edits and grey autocomplete into VS Code, powered by two small models running on your own laptop. - 13
An AI pair programmer in your terminal that edits your code and commits it, using a free local modelFix a bug by typing one sentence, watch the changes appear, and throw away anything you don't like with `/undo`. Aider edits your project and saves every change as a git commit, using a model on your own laptop, so your code never leaves your machine. - 14
Run a private AI chatbot in a browser tab, with one clickClick a link, wait a minute, and a real AI model is running on your laptop's graphics chip inside Chrome or Edge. Nothing to install, no account, and once it has loaded you can switch the Wi-Fi off and keep chatting. - 15
Build your own "chat with my files" app in Python, step by step"What's the notice period in my tenancy agreement?" Build the app that answers questions like that from your own files, quotes its sources and says when it doesn't know. Every model runs on your own computer, so nothing gets uploaded. - 16
Give your offline AI real tools: live weather, your own notes and a calculatorOut of the box, the model on your laptop can't look anything up and gets sums wrong. Tool calling fixes that: you write small Python functions, the model asks to use them, and your code decides whether it can. Build a working agent with three tools and the guard rails that keep it in line. - 17
Build your own Perplexity that has to prove every fact, on your laptopAsk a question, walk away, come back to a Markdown report with numbered citations, a sources list and a fact-check section. A local model plans the research, Crawl4AI reads the pages politely, and plain code catches the model when it makes things up. - 18
Build your own MCP server so Claude can use your toolsMCP is the plug that lets an AI app use your own code. Write a small Python server that lets Claude search your notes, add to your to-do list and check the weather, then plug it into Claude Desktop and Claude Code. It can only touch the one folder you choose. - 19
Talk into your phone, get a tidy to-do list on your PCRecord a thought on your phone. A few minutes later it's a titled, tagged note on your PC with a 3-line summary and a to-do list, made by Whisper and a local AI model. No cloud service, no subscription, and your voice never leaves your own devices. - 20
Rent a cloud GPU by the hour to run models your laptop can'tYour laptop tops out at small models. For about 60 cents an hour, rent a 48GB data-centre graphics card, chat with a model several times bigger through an encrypted tunnel, then switch it off, with a timer that does it for you if you forget. - 21
Build a Telegram bot powered by AIMessage your own AI from your phone, wherever you are. You'll build a Telegram bot in Python that answers with a model on your PC (free and private) or an API model, talks only to you, remembers each conversation, respects rate limits and keeps running on its own. - 22
Fine-tune a small AI on your own writing and run it in OllamaPrompting can describe your style. Fine-tuning bakes it in. Turn a folder of your own writing into a training set, run a real LoRA fine-tune on your laptop's CPU to prove the whole pipeline, then train a 4B model with Unsloth on a free GPU and chat with it offline in Ollama. - 23
Build an AI that uses your browser, with guardrails it can't talk its way pastGive an AI a real browser and it can compare prices, read pages and fill in forms for you. It will also follow instructions hidden in a web page if you let it. Build a small browser agent that runs on a local model, watch it get prompt-injected and stopped by code, then plug a real browser into Claude with Playwright MCP. - 24
Make AI music on your own PC, songs with vocals includedWrite a caption and some lyrics, and an open model sings them back to you as a finished song, on your own computer, with no credits to buy. ACE-Step 1.5 makes full songs with vocals under an MIT licence; MusicGen makes quick instrumentals. Learn to write captions that steer the style, split the result into stems, and know which licence lets you release what you make. - 25
Make AI video clips with Wan 2.2 on a rented GPUOpen video models now make five-second 720p clips that look like stock footage, and their licences let you use what you make. They need a big graphics card, so rent one by the hour: ComfyUI and Wan 2.2 on a pod, driven from your laptop through an SSH tunnel, with a batch script that queues a night of prompts and a calculator that tells you what each usable clip really cost. - 26
Turn invoices, receipts and forms into clean JSON with a local AIA folder of PDFs and phone photos goes in; validated JSON and a spreadsheet come out, with every document that doesn't add up flagged for a human. Ollama's schema-constrained output makes the model fill in your exact fields, Pydantic and a few lines of business rules catch what it gets wrong, and a scorer tells you how accurate it really is. Nothing leaves your PC. - 27
Talk to your AI out loud, in real time, fully offlineSpeak, pause, and a local AI answers in a natural voice; talk over it and it stops to listen. Build the whole voice loop yourself in one Python file: pause detection, Whisper, a streaming LLM, sentence-by-sentence speech and interruptions. Then tune it from 3.9 seconds to 1.7 seconds on a laptop with no graphics card. - 28
Give your webcam a brain: local AI that reads, counts and reactsA 1.3GB vision model on your laptop can read a parcel label, spot a fault on a display and tell you whether a car park is busy, in about four seconds, with nothing uploaded. Learn what small vision models are good and bad at with a real benchmark, fix their worst habit (counting), then set a watcher on your webcam or screen that saves a snapshot and alerts you when something you care about happens. - 29
ollamaBuild an AI team that researches, writes and reviews, on your own PCOne AI asked to do everything cuts corners. Three with separate jobs check each other's work: a researcher that only collects facts, a writer that only uses those facts, and a reviewer that hunts for anything made up. Wire them together with LangGraph, give them a search tool, memory and a revision limit, and measure every token, so you know what the team costs before you scale it. - 30
ollamaChoose the right AI model for your job, with your own testLeaderboards tell you which model is best at someone else's exam. Build a small test from your own work, run any mix of local and API models through it, and get one table: how often each is right, how fast it answers and what 1,000 answers cost. Add an AI judge for the open-ended answers, and catch it changing its mind when you swap the order. - 31
ollamaAn AI that reads your inbox and drafts replies in your voice, and never sends a thingEvery morning: a one-page digest of what came in and what needs you, plus replies already drafted in the way you actually write, waiting in your Drafts folder for a final read. One Python script, a local model, rules before AI, and a review step that's yours alone. It opens your inbox read-only, so it can't send, delete or even mark anything as read.
TRACK 02
Make money with AI
Turn free tools into a service local businesses pay for.
- 37 missions
- 89h 39m
- Beginner to Intermediate
- 01
Business | Website | Phone | Google reviews | On page 1? | Running ads? | Biggest problem you spotted | Company type (Ltd / sole trader) | Contact by | StatusFind local businesses that just told the internet they need youA job ad for a part-time marketer, a Google listing pointing at a dead website, a 5-star review that says "hard to find your number". Learn to spot the public signs that a business needs help this month, and build a list of 20 worth pitching with one spreadsheet and no paid tools. - 02
See any business's Google traffic, keywords and rivals for about 7 centsRun your own copy of OpenSEO, an open-source Semrush. Set it up once (about 30 minutes), then audit any local business's keywords, traffic, backlinks and AI visibility in 10 minutes, for around 7 cents a business instead of a monthly subscription. - 03
Turn the audit into a pitch deck in 5 minutesClaude writes it, Gamma designs it. You get a sharp, data-first deck that shows a business owner the money they're leaving on the table. - 04
openseoPrice it, pitch it, get paidYou've got an audit and a deck. Now turn them into money: contact the owner the legal way, run a 15-minute call that sells by showing, walk away with a real price in pounds, and get paid properly. - 05
Know before your clients do when their website goes downA bakery's website can be "up" all weekend while its Order online button has vanished. Uptime Kuma catches that, warns you before certificates and domains run out, and gives each client a status page. It's free, and it's a simple monthly service to sell. - 06
Build a fast website for a local business in an afternoon, with no monthly feesHugo turns a few text files into a finished website: prices, opening hours, a click-to-call button and a photo gallery. It scores 100 on all four Lighthouse checks, loads in under a second on a phone, and it can be hosted for free. Build it once, then reuse it for every client. - 07
Make it legal: tax, invoices and GDPR for your first clients (UK)Make £20,000 profit in your first year and your first January tax bill is not £1,931.80. It's £2,897.70, because HMRC wants half of next year's up front. This is the afternoon that stops surprises like that: the right setup, HMRC dates, records, invoices, late payers and UK GDPR, with every rule linked to its official source. - 08
Get paid on the 1st by the same clients: turn one-off jobs into monthly retainersOne-off jobs mean starting every month at zero. A retainer means the same client pays you again on the 1st, so each new client adds to last month instead of replacing it. Pick the model, scope it, price it, collect it automatically and keep clients for years. - 09
One offer, one niche, five clients: the 12-week plan for an AI service businessPick one offer, send 50 personal messages a week, aim for five clients in 12 weeks. The people who get paid chose one offer for one kind of customer and kept turning up. This is the hub of the money section: the offer, the pitch, the 12-week plan and the scoreboard, for 8 to 10 hours a week. - 10
Sell your n8n workflows: package a tested template as a product people pay forn8n's gallery holds 1,311 paid templates from 282 creators, at a median price of $20. A tested workflow with instructions a stranger can follow in 15 minutes earns its place among them. Nobody publishes sales counts, so plan around the work: a clean file, a setup guide, a demo, a price, and the setup jobs each sale brings in. - 11
Make and sell templates people actually buy: tax trackers, invoices and plannersPeople pay a few pounds for a template that saves them an evening and doesn't break. The sharpest version is selling into a law change, as Making Tax Digital pulls UK sole traders and landlords into quarterly digital records. Find what already sells, build a template that works in Excel, Word and Google Sheets, and list it within the UK rules. - 12
Find UK companies the day they register, and sell them what they need firstAround two thousand new companies join the Companies House register every day, and some are cafes, salons and builders with no website yet. A short Python script turns the free register into tomorrow morning's list. Add the UK contact rules, a fixed-price launch offer and letters that don't look like scams, and you have a lead source most web designers never touch. - 13
Win small UK public sector contracts with AI-drafted bidsCouncils, NHS trusts, universities and schools buy small jobs all year, and a one-person business is allowed to bid. In a two-week sample of award notices for this kind of work, more than half the competitions got **5 bids or fewer**. Turn the free government feed into a scored shortlist, draft answers only from work you've really done, and price bids worth winning. - 14
Turn new planning approvals into paid leads for builders and the specialists they need firstEvery week, councils publish approvals for shopfronts, extensions and lofts, each with conditions that need a specialist before work can start. Nobody sells those firms a list. Build a script that turns a week of official open data into a scored digest, and sell it as a subscription firms can say yes to on one call. - 15
Build a Shopify app and earn monthly from the App StoreMillions of shops run on Shopify, each with small problems that cost them sales, and Shopify keeps 0% of your first $1M in lifetime app revenue. Start with a free script that audits a real store's catalogue in ten minutes, and sell that audit while you build. Finding a problem merchants will pay to fix is most of the work, so this mission puts it first. - 16
Launch a micro-SaaS on UK open data and charge monthlyIn ten minutes you'll have every food business in your council rated 0, 1 or 2 for hygiene, pulled live from a free government API: a list that sells for £19 a month. Then build the alert engine, Stripe subscriptions and a £5 server. The code is the easy half; this mission also covers the weeks of outreach that land the first customer. - 17
Find the public data nobody has cleaned, and sell it as a paid APIGOV.UK answers "when are the school holidays?" with a link to 200 council websites, yet a free energy dataset flags every school holiday, per council, per day. Turn that into a paid API on real official data, with licences, Stripe tiers, hosting and the real maths done properly. Code is a third of the job; keeping the data right and finding buyers is the rest. - 18
Start a paid newsletter people actually pay forA newsletter is the one audience you own: no algorithm decides who sees it, and a reader who pays is worth more than a thousand who scroll past. The work is writing something useful, every week, for longer than you want to. This mission gets you from a blank page to a real newsletter with a paid offer and a plan for the first paying subscribers, using the same free tools and honest numbers the good ones run on. - 19
Sell print-on-demand products with AI designsMake a design with an AI image tool, put it on a t-shirt, a mug or a poster, and list it for sale without buying a single item of stock. The printing, packing and posting happen only after someone buys, so your money stays in your pocket until a sale is made. The skill is not drawing. It is picking a niche, making clean print-ready art, pricing so the maths works, and staying the right side of trademark law. Most designs sell nothing, so this is a numbers game you win by listing a lot and keeping what works. - 20
Run a review-reply service for local businesses with AIEvery local business collects reviews and almost none reply to them well. Offer owners a done-for-you service that drafts thoughtful, on-brand replies to their Google and Trustpilot reviews, saving them hours and showing every future customer a business that listens. A tested tool that runs a small AI model on your own PC comes with it, along with a reply playbook, pricing and the one line that keeps you the right side of UK law: you reply to genuine reviews only, and never write them. - 21
Run a compliant cold-email lead-gen serviceCold email still books meetings for business-to-business sellers, and most of the people doing it are breaking UK law without knowing. Learn to run it as a service, for clients or for yourself: a clean targeted list from public sources, a sequence that gets replies, mail that lands in the inbox, and results you can prove. The rules come first, because this niche is a legal minefield and getting it wrong is how you earn a fine instead of a fee. - 22
Buy, refurbish and resell laptops for profitUsed business laptops are cheap because most people cannot tell a worn-out one from a bargain. You can learn to in an afternoon. This is the full workflow: source cheap stock, test it with free tools and the real Windows commands, wipe the previous owner's data properly, price from real comps with a calculator that gives a straight go or no-go, then list and sell within the law. Margins are thin and it is work, so the numbers come first. - 23
Build a no-show-killing reminder tool and sell it to local businessesA salon with eight empty chairs a month from no-shows is losing real money every week, and most send no reminders at all. Build a working tool that reads their bookings, texts and emails each customer the day before and two hours ahead, and honours every opt-out. Then sell it, as a monthly service or a managed setup, to the clinics, barbers and garages near you. The tool, a sample booking sheet, the message templates, a privacy notice and the UK rules in plain English come with it. - 24
Turn what you know into a paid mini-courseYou already know something other people would pay a few pounds to learn faster. A mini-course packages that into a handful of short lessons someone can finish in an afternoon. Plan it with a tool that drafts the outline for you, record it on the gear you own, pick a platform on the numbers, and launch to a few real buyers. A small, focused course beats a bloated one, and finding buyers is the hard part, not the recording. - 25
Run a receipts-to-spreadsheet service for small businessesEvery sole trader has a drawer, a shoebox or a photo roll full of receipts, and a bank statement that doesn't explain itself. On the eight practice receipts in this mission, free OCR read six perfectly and flagged the two it couldn't (a faint photo and a handwritten note) instead of guessing. Turn that pile into a clean, categorised spreadsheet their accountant can drop straight in, each month, for a fixed fee. - 26
Build a paid community people stay inA paid community is recurring revenue: members pay every month for a focused room where they get something they would miss. It is also the business most people underestimate, because it only works if you keep showing up. Pick a niche, choose a platform, design tiers, onboard members well, and watch the one number that quietly caps everything: churn. - 27
Run an AI-assisted social media service for local businessesMost small business owners know they should post more, and run out of time by Tuesday. Offer them a done-for-you service: a month of on-brand posts, planned, written and scheduled, with a tested tool that drafts the whole calendar from a short brief in minutes. You can produce your first sample month tonight, before you have a single client. - 28
Get local businesses named by ChatGPT and Google's AI answersAsk ChatGPT for the best boiler engineer in your town and it names two or three firms. Every other firm is invisible to that customer. Measure it with a fixed set of prompts across five assistants, find out why rivals get named, fix what's fixable and sell the monthly number as a retainer. The first screenshot you send an owner does the selling for you. - 29
Furnish empty rooms for estate agents with AI virtual stagingAn empty lounge on Rightmove looks smaller, colder and harder to imagine living in, and every branch has a few of them live right now. Stage those rooms with an AI tool for pennies, check every image like a pro, label it properly and deliver by tomorrow 5pm. Agents can buy the software themselves; what they pay you for is the turnaround, the quality check and never having to think about it. - 30
Run paid half-day AI workshops for small teamsEvery office manager in the country has heard "we should be doing something with AI" and has nobody to ask. Walk in for one morning with a tested run sheet, live demos in their own sector and handouts they keep, and walk out with a team that's used AI on its real work. Then the 30-day check-in turns one booking into coaching. - 31
Write grant applications for charities, with AI as the drafting engineUK funders publish who they gave money to, how much and when, as free open data. Feed in the charity numbers of a few groups like your client and a script tells you exactly who funds that kind of work. Then build an evidence bank, let AI draft only from it, and charge a fair day rate for applications assessors enjoy reading. - 32
Publish beautiful modern editions of public domain booksEvery year a new batch of books falls out of copyright, and most of them are sold as ugly, unedited reprints. Choose a work that's free where you sell it, write the introduction and notes a modern reader needs, typeset it properly with Typst, and publish it on Amazon KDP and IngramSpark as part of a series. AI is your research assistant and editor; the scholarship and the taste are yours. - 33
Fix small sellers' Amazon listings: audits, rewrites and a monthly retainerThousands of small brands sell good products on Amazon behind titles that break the rules, bullets that skip the facts buyers need, and reviews that say "smaller than expected". Run one script to score a listing, find the missing keywords and turn review complaints into fixes, then rewrite it from real product facts. A free before/after on the seller's own listing is the pitch. - 34
Build email flows that make Shopify stores moneySign up to a small online shop's newsletter, leave something in the basket, and watch what arrives. Usually one email, sometimes none. Build the four automatic flows every store needs (welcome, abandoned checkout, post-purchase, win-back), write them with AI from the brand's own facts, and prove what they add with a holdout test instead of the platform's flattering numbers. - 35
Find, value and resell expiring domain namesEvery day thousands of domain names drop because their owners didn't renew. A few are short, brandable names a new business would happily pay for. Score a drop list with a script, throw out anything near a trade mark, value the survivors from real past sales, buy at registration price and list them where buyers look. It's a slow, patient portfolio game, and the maths shows exactly how slow. - 36
Get paid to train AI: RLHF, evaluation and coding tasks from the UKAI labs pay people to rate chatbot answers, write ideal responses, catch reasoning errors and review code, and the platforms that run this work recruit in the UK. Pick the real platforms, pass the one-shot assessments, track your true hourly rate with unpaid time included, and move into the coding, maths, language or professional projects that pay several times the generalist rate. - 37
Sell AI usage policies, staff training and a risk register to small firmsIn many small firms, staff already use ChatGPT or Copilot, some on personal accounts, some with client details pasted in, and nobody has written down what's allowed. Sell a pack that fixes it in two weeks: a plain-English policy built from one intake file, a one-page "can I paste this?" decision tree, an approved tools list, a scored risk register, a one-hour staff session and a sign-off sheet. Then review it every year.
TRACK 03
Go dark
Fewer trackers, fewer data brokers, a locked-down browser.
- 17 missions
- 9h 34m
- Beginner to Advanced
- 01
Find out if your password has leaked, without sending it anywhereLiverpool1 has turned up in data breaches 268,743 times, and attackers try leaked passwords first. Check yours in 10 seconds with a trick that never sends your password over the internet. - 02
Delete your password manager subscription: a free vault that never touches the cloudImport your browser's saved passwords and KeePassXC's Health Check lists every weak and reused one in seconds. Then fix them, fill them in from your browser, and carry the same vault on your phone. Everything lives in one encrypted file you own: no account, no subscription, no company server holding your logins. - 03
Host your own password manager with VaultwardenRun your own password manager sync server on your own machine, so the official Bitwarden apps on all your devices sync through a server you control instead of a company's cloud. Unlike the KeePass course, which keeps a single offline database file, this is a self-hosted sync server: you own the data, and you also take on the backups, updates and safe remote access that come with it. - 04
The Mirror's homepage makes 158 requests. Block the trackers in five minutesNews homepages quietly load dozens of ad and tracking scripts before you've read a word. Install the right uBlock Origin for your browser, prove it works, then cover your phone too. Free, no account. - 05
See which servers and countries your PC talks toAn ordinary Windows laptop talked to servers in 11 countries in 6 minutes, and one of them was a background "PC optimiser" you may not want. TapMap draws every connection your apps make on a live world map. Find out what's normal, and what to do when something looks odd. Free, no account, and it all runs on your own machine. - 06
Strip the hidden location out of your photos before you send themMost phone photos carry where they were taken, which phone took them and when. ExifTool shows you all of it, then removes it in one command, for one photo or a whole folder, without touching a single pixel. - 07
Send a 2GB video without WeTransfer, using a three-word codeVideos too big for email or WhatsApp, a whole folder of photos: croc sends them straight to another computer, end-to-end encrypted. The other person opens a link, with nothing to install. No account, nothing parked in the cloud, and it resumes if the connection drops. - 08
Get your home address off people-search sitesType your name into 192.com and see what a stranger sees: your address, who you live with, sometimes your age. This mission gets you off it, and off the copy of the electoral register those sites buy, in about 30 minutes, for free, using official routes. US readers get the big American brokers too. - 09
Stop uploading your documents to random PDF websitesA black box drawn over your account number often hides nothing: the text is still underneath, one copy and paste away. Stirling PDF redacts properly, and also merges, splits, compresses, password-protects and OCRs PDFs, all running on your own PC. No account, no installer, and your files never leave your computer. - 10
Encrypted backups you can actually restore, even after ransomware or a dead laptoprestic copies your folders to an external drive, encrypted, and keeps a history of snapshots you can step back to. You'll run a ransomware drill: scramble a folder, back up the rubbish on purpose, and still get every file back byte for byte. - 11
AirDrop for every device: send files between Windows, Android, iPhone and Mac with no cloudLocalSend moves photos, videos and documents straight across your own Wi-Fi, from any phone or computer to any other. No account, no upload to anyone's server, full-quality originals, and nothing lands on your device until you press Accept. - 12
Scan every app you built with AI for leaked API keys, before someone else doesAI coding tools love pasting API keys straight into your code, and deleting them later doesn't delete them from Git. Gitleaks digs through every old version of your project for OpenAI, Stripe, AWS and GitHub keys, and blocks new ones before they're committed. Free, open source, and it runs on your own PC. - 13
Your old gamer tag is still out there: find every account and clean it upThat username you made up at 14 is probably still attached to accounts you forgot existed. Sherlock checks a username on about 400 websites in around 30 seconds. Run it on yourself, find the forgotten accounts, then close or lock them down before anyone else goes looking. - 14
Block ads and trackers on every device in your house, including the smart TVOne news homepage contacts 31 servers before you've read a word, and 18 of them are ads and trackers. AdGuard Home blocks those for every device in the house: phones, laptops, consoles and the smart TV you can't install anything on. Test-drive it on your PC in 15 minutes, then move it to a Raspberry Pi or an old laptop and point your router at it. - 15
Lock any file with one command, so only you (or the person you choose) can open itSend a message over WhatsApp or email that only one person can read. Lock your passport scans before they go anywhere near the cloud. age turns any file into scrambled bytes that only your passphrase or your key can open. Free, tiny, no account, and it runs on Windows, Mac and Linux. - 16
Harden Firefox until websites think you're in Iceland, without touching your normal browserBefore: websites read your real CPU, graphics card, screen and time zone. After: 4 generic cores, a graphics card called "Mozilla" and a clock in Reykjavik. arkenfox is the settings file privacy pros use to harden Firefox, about 150 changes in one go, set up here in its own profile so your everyday browser stays exactly as it is. - 17
Run your own VPN with WireGuardA VPN is one of the most oversold tools on the internet, so this mission does two things at once. First it is straight with you about what a VPN actually does (it encrypts the hop to a server you choose and changes your apparent address to that server's; it does not make you anonymous). Then it has you build a real one: you make the keys, write the config, bring up a genuine WireGuard tunnel between two network namespaces on your own machine, and watch it handshake and carry real encrypted traffic. By the end you understand VPNs well enough to see through the adverts, and you can put the same server on a cheap VPS or your home router so you reach your own network safely from untrusted Wi-Fi. Nothing here is for hiding wrongdoing: it is your own encrypted tunnel to a server you control.
TRACK 04
See everything
Watch the world's public data like an analyst: flights, fires, markets and news on one screen.
- 5 missions
- 1h 50m
- Beginner to Intermediate
- 01
A spy-satellite view of the world, in your browserGod's Eye View turns public data into a 3D globe that looks like a classified cockpit: live aircraft with their call signs, thermal and night-vision filters, and any city on Earth. No account, no keys. - 02
Your own situation room: conflicts, nuclear sites and live TV on one free mapWorld Monitor puts conflict zones, military bases, nuclear sites, live news channels, markets and disasters on one live map, in 2D or 3D. It runs on your own PC in about 15 minutes, with no account and no API keys. - 03
Run your own live world-watch dashboardThe preview clip shows what you'll have running in 20 minutes. Crucix pulls flights, markets, satellites, radio receivers and world news from public feeds into one map on your own PC, with no account and no API keys, and refreshes itself every 15 minutes. Then add a free NASA key to light up the fire map, and get alerts on your phone through Telegram. - 04
Satellites, earthquakes and world news on one dark mapRight-click almost anywhere on Earth and get a recent satellite photo of it. ShadowBroker puts that on a dark map with dozens of public feeds: satellites in orbit, earthquakes, volcanoes, conflict reports and breaking news. It runs on your own PC with no accounts, and two free keys fill the skies and seas with live planes and ships. - 05
Turn a satellite's radio beeps into a real photo of EarthWeather satellites beam pictures of Earth down to anyone with a radio that costs about the same as a takeaway. SatDump turns that signal into images of clouds, coastlines and storms. You start with a real recording, so you'll have your first satellite image in 15 minutes without buying anything.
TRACK 05
Reverse engineer anything
Take apart programs, code and data the way analysts and intelligence agencies do, with the free tools they actually use.
- 4 missions
- 2h 12m
- Beginner to Intermediate
- 01
Decode anything with GCHQ's CyberChefBritain's intelligence agency gave away its "cyber Swiss army knife". It runs in your browser, needs no install, and turns scrambled text, scam links and mystery files into something you can read. - 02
Get the prompt behind any GitHub project in 4 secondsPaste a GitHub link, get back the plain-English prompt someone could give Claude or Cursor to build that project. Free, in your browser, no account, about 4 seconds. - 03
Find the secret data feed behind any live website, and turn it into a spreadsheetMost live websites quietly download their data as JSON in the background. Spot that request in your browser, replay it in the terminal, and turn it into a script that saves a spreadsheet and a map. Practice target: the USGS live earthquake map, which is public, free and needs no account. - 04
Crack open a program with Ghidra, the NSA's free toolGhidra is the reverse-engineering tool the NSA built and then gave away for free. In this mission you build a tiny password-locked program yourself, then use Ghidra to read its compiled code, recover the password it was hiding, and patch the lock out entirely. Nothing illegal, nothing online: it's your program on your own machine.
TRACK 06
Ethical hacking
Learn to attack and defend like a security professional, legally, in a lab on your own machine against software built to be broken.
- 15 missions
- 13h 29m
- Intermediate to Advanced
- 01
Build a hacking lab on your own PCBefore you can learn to attack anything, you need something you are allowed to attack. In this mission you stand up a safe, isolated practice lab on your own machine: Docker running, a private network the lab lives on, and OWASP Juice Shop (a web app built on purpose to be broken into) waiting at 127.0.0.1. Nothing is online, nothing points at anyone else, and one command wipes it back to clean. This is the foundation for the whole ethical-hacking track. - 02
Web app pentesting with Burp SuiteCourse 1 built you a legal target: OWASP Juice Shop running in Docker on your own PC. Now you meet the tool professional web testers reach for first, Burp Suite. You install the free Community Edition, send your browser's traffic through it so you can watch and change it, read one real request and its reply byte for byte, change a single value in Repeater and resend it, and then solve three genuine Juice Shop challenges the proper way: logging in as the administrator through SQL injection, walking into the admin section, and reading another customer's shopping basket. Every request in this mission goes to 127.0.0.1 and nowhere else. - 03
Find bugs before attackers do: fuzzing your own codeA fuzzer is a tool that throws millions of weird, malformed inputs at a program on purpose, trying to make it crash, so you can fix the hole before anyone else finds it. In this mission you install Atheris (Google's fuzzer for Python), write a tiny parser with a bug planted in it, and watch the fuzzer discover a crashing input in a fraction of a second. Then you read the crash, shrink it to the smallest input that still breaks things, fix the code, and prove the crash is gone. This is defensive security from start to finish: you only ever attack a small program that you wrote yourself. - 04
Harden a Linux server, and prove it workedThis is the defend side of the track. You stand up a deliberately weak Linux server in a container on your own machine, measure how exposed it is with a real scanner (nmap), then apply the hardening a professional would: key-only SSH with no root login, a firewall, fail2ban, removing services you do not need, and automatic security updates. The proof is the before and after: the same scan that found three open ports and password logins returns one open port and refuses passwords once you are done. Everything happens against your own practice box in Docker, never anyone else's machine. It pairs with the Build a hacking lab and Burp lessons: those teach you to attack, this teaches you to close the doors. - 05
Read a CVE and reproduce it in your labA CVE is a published, catalogued security vulnerability. Reading the record is one skill; understanding how the flaw actually works is another, and the way professionals learn the second is to reproduce a known vulnerability safely, on their own machine, against software they installed themselves. In this mission you take a real, well-documented flaw in a popular Python library (CVE-2021-33503, a denial of service in urllib3), install the exact affected version in an isolated environment, trigger the bug for real and watch the time it wastes climb, read the few lines of code that caused it, then upgrade to the fixed version and prove the same input is now harmless. Nothing is attacked except a library on your own PC. It pairs with Build a hacking lab, which sets the one rule everything here obeys. - 06
Find the known vulnerabilities in your own Docker imagesEvery Docker image you build quietly carries hundreds of other people's packages: the operating system underneath it, the language runtime, and every library you install. Any one of them can have a published, known security flaw, and you cannot fix what you cannot see. In this mission you install Trivy, a free scanner, and point it at an old image on your own machine. You read the report it produces, learn what the severity numbers mean and which findings have a fix, then you write a small Dockerfile, scan it, rebuild it on a current base image and watch the vulnerability count fall. Everything here runs against images you built or pulled yourself. It is defensive supply-chain work, the kind that stops a known flaw reaching production, and it pairs with Build a hacking lab, which set up the Docker you will use. - 07
Nmap NSE deep dive: script your own labMost people stop at nmap -sV. The real power is the Nmap Scripting Engine (NSE): 600-odd small scripts that go past "which ports are open" and ask "what is running, how is it configured, and is any of it weak". This mission is a hands-on tour of NSE against a target built to be broken, OWASP Juice Shop, running on your own machine. You learn what the script categories mean, run default, discovery and vuln scripts against the lab box, read the output (including a textbook false positive the vuln category reports), customise a script with its own arguments, and learn which categories are safe to run and which can take a server down. Everything here points at 127.0.0.1, a practice target you start yourself. It sits alongside Build a hacking lab and Harden a Linux server: those stand up and lock down a box, this one teaches the scanner's deep end. - 08
See exactly what your computer is sending with WiresharkYour laptop is talking to the internet all the time, and most of it you never see. Wireshark lets you watch it: every connection your apps open, every name they look up, every server they reach. In one short capture on your own machine you will read a real plaintext web request word for word, watch a name lookup turn github.com into an address, and list every host you talked to. You will also see, honestly, where the view stops: most traffic now is encrypted, so you learn to read the metadata (who, where, which service) rather than the contents. Free, open source, and entirely on your own machine and network. - 09
Stop XSS and clickjacking with security headers and a CSPThis is the defend side of web security, the companion to the Burp lesson where you attacked a web app. Here you run a tiny web page on your own machine that has one deliberate hole: it reflects whatever a visitor types straight back into the HTML without escaping it. You watch an injected script run and change the page, then you add the HTTP security headers every site should send, write a Content Security Policy that blocks the injection while still allowing your own scripts, and roll it out safely with Report-Only so you never break your own site. The proof is the before and after, captured for real: the headers absent then present, the injected script running then refused, and the browser's own Content Security Policy violation message. Everything runs against a demo page you build and own, on 127.0.0.1, and nowhere else. - 10
Test your own API for the flaws attackers look forA website hides a lot behind its buttons, but the API underneath answers anyone who asks it directly, and that is where most modern breaches happen. In this mission you test a REST API for the five flaws that top the OWASP API Security Top 10: broken object-level authorisation (reading another user's data by changing an id), broken authentication (a missing or forged token that still works), excessive data exposure (an endpoint that hands back fields it should hide), missing rate limiting (a login that never slows a password-guesser down), and mass assignment (setting a field like is_admin the app never meant you to touch). You trigger each one for real with curl and httpx against two targets on your own machine: a tiny API you write yourself, deliberately broken so you can see every flaw, and OWASP Juice Shop, the intentionally vulnerable app from course 1. Then you fix every flaw in your own code and prove, with the same requests, that the holes are closed. Every request goes to 127.0.0.1 and nowhere else. - 11
Read server logs like an analyst and find the attackThis is the blue team side of the track: the work of the defender who reads what a server wrote down and spots the attack inside it. You build a tiny log lab on your own machine (no root, no Docker), run a real OpenSSH server on localhost, then knock on it with failed logins the way a brute-force bot would, so the logs fill with real attack entries you created yourself. Then you do the analyst's job on them: read the SSH auth log with journalctl, find the failed logins and the address they came from, count the attempts per source with a grep and awk pipeline until the one noisy attacker stands out from the quiet, read a web server access log for the scanners sweeping for vulnerable pages, and finally turn detection on with fail2ban and summarise the web log with GoAccess. Every log line you analyse is one your own activity produced, against your own box. It pairs with Harden a server (the defence you apply once you can see the attack) and Build a hacking lab (which sets the one rule everything here obeys). - 12
Crack your own passwords to see how strong they really areThe fastest way to understand why one password is safe and another is not is to attack them yourself. In this mission you install the two best-known password crackers, hashcat and John the Ripper, inside WSL2. Then you invent a few passwords, turn them into hashes the way a real site stores them, and point the crackers at your own hashes. A weak one falls in about a second. A long passphrase does not fall at all: the number of guesses needed is larger than the tool can even count. Everything here happens against hashes you made up on your own machine, which is the one lawful way to do this. You leave with a concrete rule for your own passwords, and the evidence behind it. - 13
Find the risk hiding in your project's dependenciesAlmost every project you build stands on other people's code: the packages you install, and the packages those packages install in turn. Any one of them can carry a published, known security flaw, or be a hostile lookalike of a package you meant to install. In this mission you build two small projects on your own machine, one npm and one Python, each with deliberately old, vulnerable dependencies. You run real audits (npm audit and pip-audit), read the advisory ids and severities they return, then upgrade and pin and watch the same audit come back clean. You learn what a lockfile is and why it matters, how to inspect a package before you install it, and how to spot a typosquat before it bites. Everything runs against your own projects. It is defensive supply-chain work, and it pairs with Read a CVE and reproduce it, which goes deep on a single flaw. - 14
Run code you do not trust in a safe sandboxSometimes you have to run the thing you are unsure about: a script from a stranger, a macro in a file a client sent, a proof-of-concept from a security advisory. This mission builds a throwaway, sealed box to run it in, a Docker container with no network, a read-only filesystem, no Linux superpowers, hard caps on processes and memory, and a non-root user, so that even if the code is hostile it cannot reach your real files, your network, or your system. You prove each wall holds by running a small, harmless probe inside that tries to break out and fails at every turn. It pairs with Analyse a file without running it (which covers looking before you leap): this is the one for when you have to leap. - 15
Inspect a suspicious file safely, without running itA file lands in your downloads or your inbox and you do not trust it. The dangerous instinct is to double-click it to 'see what it is'. The professional move is the opposite: examine it from the outside, without ever letting it run, and decide whether it is dangerous before it gets the chance to do anything. That is static analysis. In this mission you learn to read an unknown file the way an analyst does: its true type, its hash, the text and web addresses hidden inside it, its metadata, and anything a second file is hiding inside it, then look its hash up against the world's antivirus engines. You practise on the EICAR antivirus test file (a harmless standard that every scanner recognises) and on sample files you build yourself, never on real malware and never on anyone else's private files. It pairs with Build a hacking lab, which sets the one rule everything here obeys.
TRACK 07
Going ghost
Take back control of what the internet knows about you, from a clear-eyed plan to the tools that actually move the needle.
- 3 missions
- 3h
- Beginner to Intermediate
- 01
Build your personal threat model (decide what you are actually protecting)Before you buy a VPN, delete your accounts or encrypt your hard drive, answer three questions: who are you hiding from, what are you protecting, and how much effort is it worth? That short exercise is called a threat model, and it is the difference between privacy that fits your life and a pile of half-used tools that lock you out of your own accounts. This is the foundation course of the Going Ghost track. You finish it with a one-page written plan, specific to you, that tells you which of the later courses to do first and which to skip. Perfect privacy is a myth, and chasing it has real costs. This mission helps you spend your effort where it actually counts. - 02
Compartmentalise your digital life (one leak shouldn't expose everything)Most people live their whole online life in one browser signed into everything at once, which is why a single breach, nosy service or data broker can join the dots between your work, shopping, banking and public posts. Compartmentalising keeps those identities in separate boxes (browser profiles, Firefox containers and a different email per service) so a crack in one does not drain the rest; it is course 2 of the Going Ghost track, for lawful personal privacy, not for deceiving anyone, committing fraud, evading a ban you earned, or dodging a lawful investigation. - 03
Tor, explained safely (what it really does, and how to use it without fooling yourself)Tor is the most misunderstood privacy tool there is: some think it makes them invisible, some think it is only for criminals, and both are wrong. This mission gives you the real picture: what onion routing actually does (your traffic is wrapped in three layers of encryption and bounced through three relays, so no single one knows both who you are and what you are looking at), how to install the Tor Browser, and how to prove with your own eyes that your traffic genuinely leaves the Tor network. It is for lawful privacy and getting around censorship, not a route to anything criminal; pair it with your threat model and reach for Tor when your model calls for it, not by default.
TRACK 08
Creator toolkit
Free AI tools that replace the paid creator apps: subtitles, cut-outs and voiceovers, all on your own laptop.
- 13 missions
- 6h 19m
- Beginner to Intermediate
- 01
Subtitles for any video, free and offlineBig, bold, Reels-style captions burned into your video, made by OpenAI's Whisper on your own laptop. You also get an .srt for YouTube and a plain transcript. No subscription, no upload, no graphics card. - 02
Shrink, trim and reformat any video in secondsFFmpeg is the free engine inside most video apps. A handful of copy-paste commands cover what creators pay online converters for: smaller files, vertical Reels versions, clean trims, burned-in captions, level audio, speed-ups, GIFs and MP3s. - 03
Remove the background from any photo in one commandCancel the remove.bg credits and the Canva Pro trial. rembg cuts out product shots, thumbnails and profile pics on your own laptop, a whole folder at a time, and your photos never get uploaded. - 04
Make small, blurry images 4x bigger and sharper for freeA 256-pixel photo of Einstein became a sharp 1024-pixel image in 8 seconds. Upscayl does that to old family photos, tiny thumbnails and AI art on your own laptop: no upscaler credits, no uploads, and no gaming graphics card needed. - 05
Copy the text out of any screenshot, even a photo of a signGrab text off your screen in ten seconds with a tool you already have, then go further: searchable PDFs from scans with OCRmyPDF, and text from photos and screenshots with RapidOCR. Free, offline, and nothing gets uploaded. - 06
Remix a film made entirely of code, and post it as a ReelWindow Seat, Roost and the other risograph films from our reels are each one HTML file: no video, no images, just code that draws every frame. Change the colours and the fireworks, then export your version as a vertical Reel with sound. - 07
Free AI voiceovers on your laptop, no ElevenLabs billNatural-sounding narration for videos, tutorials and audiobook drafts, made by the open Kokoro model on an ordinary laptop. No subscription, no account, no graphics card. - 08
Remove the vocals from any song in 2 minutes, freeDemucs, the free AI model first built at Meta's research lab, pulls a song apart into vocals, drums, bass and everything else. Karaoke tracks, a cappellas, practice tracks and a finished karaoke video, all on your own laptop with no uploads and no credits. - 09
Cut the silences out of any talking video automaticallyStop scrubbing through your timeline hunting for dead air. auto-editor listens to your video and removes every pause in one command: a 70-second clip came out at 36 seconds in about 4 seconds of work. It's free, it runs offline, and it can hand the cuts to Premiere, DaVinci Resolve, Final Cut or Shotcut so you can fine-tune them. - 10
Record tutorials that look like a YouTuber made them, free, with OBSSet up OBS Studio, the free recorder streamers and YouTubers use, so it records sharp 1080p tutorials with your logo on top and a clean, even voice, then flip it to a vertical 1080x1920 canvas for Reels and TikTok. - 11
Turn a blurry logo into a sharp vector you can print on anything, freePlenty of small businesses only have their logo as a fuzzy social media picture, and every printer they call asks for a vector file. VTracer redraws it as an SVG: clean shapes that stay sharp on a mug, a van or a banner. Free, open source, under a second per logo, and a cleaned-up vector is a small job you can charge for. - 12
Write a web page, get a video: make Reels and explainers with HTMLHyperFrames turns one ordinary HTML file into a real MP4. Get any AI chatbot to write the file, watch it in your browser, then render vertical videos on your own laptop. Free, open source (Apache 2.0) and no account. - 13
Video too big for WhatsApp or email? Shrink it in one clickHandBrake is the free, open-source video shrinker with ready-made presets. Pick one, press Start, and a 96MB clip fits in an email at 16MB, still looking good. No account, no watermark, and nothing is uploaded.
TRACK 09
Fix your PC
Find what's slowing your computer down and eating its space, with free tools that only change what you choose.
- 1 mission
- 22m
- Beginner
TRACK 10
Off-grid & offline
Knowledge, maps and media that keep working when the internet doesn't.
- 3 missions
- 1h 20m
- Beginner to Intermediate
- 01
Carry Wikipedia on your laptop, offlineBuild a grab-and-go survival library: the best 50,000 Wikipedia articles and an emergency medicine guide on your laptop or a USB stick, searchable with no signal at all. The whole of English Wikipedia, every article in full, fits in 49GB. Free, no account, nothing to install. - 02
Your whole city's map in 20MB, working with zero signalCut your town out of a free world map into one small file, then browse it down to street level with the internet switched off. Central London comes to 20MB, small enough for a USB stick. No account, nothing to install. - 03
Download whole video channels for flights, with subtitles, in one commandPoint the free tool yt-dlp at a channel the night before you travel and it saves every new video at the size you choose, with subtitles, skipping the ones you already have. Films, talks as tiny MP3s, whole series. Step 1 shows which videos are yours to keep.
TRACK 11
Cancel your subscriptions
Run your own versions of the apps you pay for, on a computer you already own.
- 10 missions
- 5h 38m
- Beginner
- 01
Your own private Dropbox, with no cloud and no subscriptionStop paying for cloud storage. Syncthing keeps folders in sync between your laptop, desktop and phone, directly from device to device, and sends your phone photos to your PC by itself. No account, no monthly fee, and your files never sit on a company's server. - 02
Cancel Evernote: your own encrypted notes app, for £0Evernote's paid plans charge every year to sync notes you wrote. Joplin does the notebooks, tags, to-dos, search and web clipping for nothing, syncs through storage you already have, and encrypts every note before it leaves your laptop. Your old notes come with you. - 03
Cancel your budgeting app: a free, private budget on your own PCImport a month of bank transactions and watch seven simple rules sort 20 of them into the right categories. Actual Budget gives you that, plus envelope budgeting and spending reports, running on your own computer with no subscription and no account with anyone. - 04
Watch thousands of free public TV channels from around the worldiptv-org is one of the biggest projects on GitHub: a list of over 11,000 TV channels that broadcasters stream publicly for free. Paste one link into VLC and browse news, sport and culture from almost every country. - 05
Your own Netflix for the films and shows you already ownJellyfin turns a folder of video files into a Netflix-style app with posters, descriptions and "carry on watching". Free, open source, no account with anyone, and it runs on the laptop you already have. - 06
Cancel Microsoft 365: a free office suite that opens your Word and Excel filesMicrosoft 365 costs £84.99 a year for one person (£104.99 for a family). LibreOffice opens, edits and saves the same Word, Excel and PowerPoint files for £0, with formulas, tables and charts intact. It even does something Word can't: 40 Word files turned into PDFs in 35 seconds, without opening one. - 07
Your own Spotify for the music you own, on your phone at home and awayNavidrome turns a folder of music files into a Spotify-style app with album art, playlists and a player in your browser, and dozens of phone apps can stream from it. Free, open source, no account needed, and it runs from one folder on your laptop. - 08
Your own start page: news, weather, markets and Reddit on one calm page, with no algorithmGlance is one small program that builds a personal homepage from a single text file. You pick the sources, it shows the newest first, and there's no feed deciding what you see. - 09
Your own Airtable: turn a spreadsheet into a real database with forms, views and a kanban boardNocoDB turns a spreadsheet into a real database you run on your own PC: proper field types, linked tables, a kanban board, read-only share links and a free API. You'll build it with a sample cake shop's 10 orders and 5 customers. - 10
Your own ebook library: organise, convert and read your books on any device, freeYour whole book collection on your phone, in a reader that runs in the browser, served from your own laptop over Wi-Fi. calibre is the free, open-source ebook manager that does it: a tidy library with covers and series, a one-line converter for Kindle and Kobo, and a built-in web reader. No account, no subscription, and it all lives in one folder. - 11In testing
Your own Google Photos with Immich
TRACK 12
Automate the boring stuff
Free tools that type, click, watch and file things for you, so you stop doing the same job twice.
- 8 missions
- 9h 21m
- Beginner to Intermediate
- 01
Type :reply, get a whole email: free text shortcuts that work in every appEspanso turns short codes like :addr into your full address, a whole email or today's date, wherever you type. Free, open source, no account. - 02
Your own Zapier: automate boring jobs on your own PC, freen8n is a drag-and-drop automation builder, like Zapier, that runs on your own computer. You'll build a real one: every morning, the BBC tech stories that mention AI land in a spreadsheet. No account, no API keys. - 03
Get an alert the moment a price drops or a web page changesPick the price you'd pay and your phone buzzes the moment the shop hits it, or the moment a sold-out item is back. changedetection.io runs on your own computer: free, open source, no account. - 04
Build a robot that does your boring web admin while you make a teaDo the job once by hand: Playwright watches you log in, search and download, writes it down as a script, then repeats it for a whole list on its own. Free and open source, from Microsoft. You practise on a fake supplier portal first, so nothing real is touched. - 05
Run n8n 24/7 on a £5 server, with HTTPSYour n8n on your PC stops when the laptop sleeps. Move it to a small rented server with your own address, a real HTTPS padlock, webhooks other apps can reach, and nightly backups. Docker, Caddy and Ubuntu, set up properly, one step at a time. - 06
Sort, label and summarise your email with AIBuild an inbox assistant in n8n that sorts every new email into five categories, labels it in Gmail, and sends you one short briefing each morning. The AI can run on your own laptop with Ollama, so your email never leaves it. It never sends a reply, never deletes anything, and you can switch it off from your phone. - 07
Get a ping on your phone the moment your wishlist drops in priceBuild a small Python price tracker that checks your wishlist once a day, reads prices the way Google does, keeps a spreadsheet history and buzzes your phone when something gets cheaper. You build it on a practice shop on your own PC, then point it at real shops politely and legally. - 08
Make your scripts run themselves: nightly backups and free site-down alertsA script you have to remember to run is a script you'll forget. By the end, your files back themselves up every night, GitHub checks your website every morning while your PC is off, and your phone buzzes the moment either one fails. Task Scheduler, cron, launchd and GitHub Actions, each with a clean off switch.
TRACK 13
Build things with code
Make videos, games, apps and animations with code, even if you've never programmed.
- 18 missions
- 18h 1m
- Beginner to Intermediate
- 01
Make a Reel-ready animated intro with codeBuild one vertical title card in React, then turn a simple list into a folder of finished Reel intros with one command. Remotion films the code into real MP4s on your own laptop. Free for individuals and small teams. - 02
Make 3Blue1Brown-style animations with a few lines of PythonVertical maths and data clips for Reels, Shorts and TikTok, written in a few lines of Python. Manim is the free animation engine behind the 3Blue1Brown videos: shapes that morph, charts that grow, a dot riding a graph, and a 15-second explainer you fill in with your own idea. - 03
Turn a Python script into a web app you can shareStreamlit turns a short Python file into a real web page with sliders, dropdowns, file uploads, charts and download buttons. You'll have a working quote calculator in your browser within ten minutes, then build a sales dashboard and put it online for free. No HTML, no JavaScript. - 04
Where did my money go? Explore your spending in a Python notebook that turns into an appSix months of card payments, one slider: drag it and every total, chart and table updates at once. marimo is the free Python notebook that does it. Explore your own bank export with a dropdown, a chart and SQL, then turn the same file into an app and a web page you can send. - 05
Build a live space-station tracker (and more) from free APIsLearn to read the famous public-apis list, then build one HTML file that shows the space station moving on a live map, London weather, the next UK bank holidays and a joke. No keys, no sign-ups, nothing to install. - 06
Script a terminal demo once, render it as a GIF or a ReelVHS reads a short text file, types your commands into a hidden terminal and films it as a GIF, MP4 or WebM. No screen recorder, no typos, no retakes. When your project changes, edit one line and render it again. - 07
Find the move that lost your last chess gameStockfish, the world's strongest chess engine, is free and runs on your own laptop. You'll talk to it directly, then run a short Python script that marks every inaccuracy, mistake and blunder in a game and shows you the move you should have played. - 08
Put your website online free with GitHub Pages and Cloudflare PagesA website is just a folder of files, and two big companies will host that folder for nothing. Your site is live on GitHub Pages in about 35 minutes, then Cloudflare Pages adds preview addresses and one-click rollbacks. - 09
Point your own domain at your site, with HTTPSSwap your free github.io or pages.dev address for yourname.co.uk, padlock and all. Buy the name at a fair price, add the few records your host asks for, and check every step with the included checker, without breaking anyone's email. - 10
Build your own API in Python: a private link-saver for every deviceSpot a good link on your phone, save it with one tap, find it later on your laptop. Build linkstash, your own private link-saver with tags, search and an API key, using FastAPI and SQLite. It writes its own documentation, ships with tests, and finishes online behind HTTPS for your phone and automations to use. - 11
Ship your own command-line app that anyone installs with one lineBuild `tidyup`, a small Python tool that sorts a messy Downloads folder, shows you the plan first with `--dry-run`, and can undo itself. Then install it like a real program with one uv command and share it from GitHub in one line. Publishing to PyPI is an optional last step. - 12
Build your first Chrome extensionHighlight anything on the web, right-click, and it's saved to your own notes, ready to copy as Markdown. Build it from scratch in plain JavaScript, with real Manifest V3 code and no build tools, for Chrome and Edge. Then grow it into a page summariser that runs on a free local AI. - 13
Docker without fear: run any self-hosted app in a containerMost self-hosted apps say "just run this Docker command". By step 3 you'll have a website monitor running on your own PC from one line, then a private budgeting app next to it. Along the way, learn what images, containers, volumes and ports are, and the habits that keep your data safe. - 14
Beautiful invoices and CVs from plain text, and a CV service you can sellTypst turns a small text file into a polished PDF in a fraction of a second. Set your invoice up once and every new one is a pasted message and one command. Then use the same trick to sell one-page CVs and branded invoice templates, free and offline. - 15
Turn plain text into branded Word files, PDFs and ebooks clients will pay forWrite a proposal once and get a branded Word file, a PDF, a web page and an ebook from it, each in under a second. Then turn a local business's blog into a free ebook they swap for email sign-ups, and charge for it. Pandoc is free, open source and needs no account. - 16
Pitch decks from plain text: your own template, a deck per prospect, a PDF by lunchWrite your pitch as one text file. Slidev turns it into a clean deck with click animations, a price table, a diagram and speaker notes, then exports a PDF you can email. It's the free, offline version of the Gamma lesson, with a template that's yours, so one spreadsheet of prospects becomes one personalised deck each. - 17
Analyse a million-row spreadsheet in seconds, when Excel gives upOpen a 2-million-row sales file in Excel and it quietly stops at row 1,048,576: 951,425 sales gone, and every total wrong. DuckDB reads the whole file and answers in under half a second, then hands the answers back as a small Excel sheet. By the end you'll have a monthly report you can sell to any business sitting on exports it never looks at. - 18
Find every slow page on a website, fix it, and prove it: 28 seconds down to 2A bakery home page took 28.2 seconds to show its main photo on a phone; two fixes later it took 2.1. Unlighthouse runs Google's Lighthouse test on every page of a site at once and puts the slow pages, missing titles and broken bits in one report. Turn that into a fix list a business will pay for, and a monthly check they'll keep paying for.
TRACK 14
Services you can sell
Turn the free tools in the Underground into services small businesses and creators already pay for, one clear offer at a time.
- 22 missions
- 61h
- Beginner to Intermediate
- 01
Sell a website care plan: monitoring, speed checks and a monthly reportSmall business websites break quietly: a certificate runs out, an update kills the booking form, nobody renews the domain. Be the person who spots it first, fixes it and sends a one-page report every month, for about an hour per site. Two client sources most people miss come with it: owners whose web designer has vanished, and CQC-registered practices whose sites quietly break the rules. - 02
n8nAutomate a small business's boring admin with n8n, and get paid for itA small business advertising for a part-time admin assistant has just told you its budget: about £9,900 a year. Get there first with "before you hire, I can take half that list off you for a fixed £350". Six ready-made n8n workflows, a personalised demo, a call script and the licence rules for selling n8n work come with it. - 03
Spot a local business with a broken website, and quote the fix properlyShow a café owner their own website, fixed, on your phone, in ten minutes, without touching their real site. Then learn the 12 problems almost every small business site has, check each in about two minutes with free tools, and turn them into a fixed-price quote the owner can say yes to. - 04
Turn a messy podcast recording into a clean, subtitled episode with one script, then sell itA rough recording goes in at -27 LUFS with the guest buried under the host, and one script later it comes out levelled at -16 LUFS, pauses cut, with subtitles and a transcript. Do that first, on your own PC. Then turn it into a service: who pays, three packages, about an hour of hands-on work per 30-minute episode once you're practised, and clients won within UK marketing rules. - 05
Get a plumber into the Google map pack, legally: their profile, reviews and website detailsSearch "plumber near me" and three businesses get the map; everyone else might as well be closed. Fixing a neglected Google profile is checklist work you can learn in a month of practice, and the owner sees the result on their own phone. Your first door-opener takes ten minutes: a calm reply to the 1-star review they never answered. - 06
Add a chatbot to a client's website that answers their FAQsEvery small business answers the same ten questions all day: are you open, how much, is there parking. Build a chat assistant from a business's own website in half an hour, show it to the owner before you've asked for a penny, then sell them the upkeep. It answers only from the approved FAQ, hands everything else to a person, and costs about a pound a month to run. - 07
Find businesses paying Google for clicks to a slow website, and fix itPlenty of local businesses pay for Google Ads that land on a page taking seven seconds to appear on a phone, and they have no idea, because it looks fine on their laptop. Find them with free public tools, apply the 10 fixes that solve most small-business slowness, and hand over a before and after report that sells the next job. - 08
Make a messy online shop look like a real brand: product photo clean-upsSmall online sellers shoot their products on a kitchen table, and their collection page looks like a car boot sale. Turn it into a grid of clean, matching photos in an afternoon, show the owner their own shop transformed, and sell it as a monthly new-stock plan. Free tools do the cutting; you bring the batch workflow and the quality check that catches what the tools miss. - 09
Subtitles and translations as a service, done on your own laptopCouncils, universities and NHS trusts publish accessibility statements that admit, in writing, "some of our videos do not have captions". A free script counts exactly how many on their YouTube channel, so you walk in with that number and a price. Then deliver with a laptop workflow that turns a 10-minute video into about an hour of hands-on work. - 10
Sell the logo file every sign maker asks for (plus thumbnail fixes as an add-on)Plenty of high street businesses have only a blurry JPEG of their logo, right when the sign maker asks for "a vector file". Rebuild it and hand over a proper 32-file logo pack, made by one command from the kit. Find owners at exactly the right moment through council sign applications and local printers, with YouTube thumbnail fixes as an add-on. - 11
Turn one long video into ten short clips for a creatorEveryone who records long videos knows they should be posting short clips, and most never get round to it. Turn an hour of video into ten captioned vertical clips in about two hours of your time, with a tested routine and three packages to sell. You can make your first hook map for a real creator tonight, before you've cut a single clip. - 12
Sell reconciled bank statements to accountants: piles of PDFs in, checked spreadsheets outOn the practice scan in this mission, every total matched, 15 rows were wrong, and only a running-balance check caught it. Somewhere in every accountancy practice, someone is typing a year of bank statements into Excel by hand. Do it faster with free tools on your own laptop, and better, because every batch comes back reconciled and proven. - 13
Tell a café what actually sells: a £250 one-page report from their messy spreadsheetThe practice till export in this mission has 198 rows pasted in twice, quietly inflating every total the owner has ever looked at. Owners don't want a cleaner spreadsheet; they want to know what sells, what doesn't and when to put more staff on. Clean the data with a tested DuckDB or Python script, answer those questions in a one-page report, and sell it with a monthly refresh. - 14
Pitch decks for small businesses, made from plain textEvery week, small businesses and charities stand up in front of a buyer, a funding panel or an investor with 30 cluttered slides, and the dates are public in advance. Turn their mess into a clear 10-slide deck, a one-page summary and speaker notes in a week. Find them before their deadline, run the interview, and build every deck on a reusable Slidev theme. - 15
Sell accessibility audits now the European Accessibility Act is lawThe free scanners everyone uses can't see the barrier that costs shops the most: a keyboard user who can't reach Add to basket. Since 28 June 2025, EU law has required many online shops selling to EU consumers, UK ones included, to be usable by disabled customers. Find what the scanners miss with a keyboard and a free screen reader, and sell the report at a fixed price. - 16
Build MCP servers for software companies: the new integration gigPlenty of small software companies have a good API, customers asking for "it in Claude", and no time to build or maintain an MCP server. Build the official one for their API, hand it over tested and documented, and look after it as the spec moves. Few people can do both the build and the security properly, which is exactly why it pays. - 17
Sell an AI phone receptionist to tradespeople who miss calls"Ring this number and pretend your boiler's broken." That one demo sells it. Set up a voice agent that answers when the trade can't, takes the job and texts it over within a minute, and sends gas leaks, carbon monoxide and danger to life straight to the right emergency line, every time. - 18
Dub YouTube channels into new languages with multi-language audioPress play on the demo in step 1: a 48-second English how-to, then the same video speaking Spanish, timed to the picture with free tools on a laptop. YouTube's free auto-dubs can't be edited when they get things wrong, and a reviewed, well-timed dub fills that gap. Bilingual? You already have the hardest part. - 19
Google's new image rule lands in January: get paid to fix small shops' Shopping feedsFrom 31 January 2027 Google wants every product photo at least 500 x 500 pixels, on top of the bad barcodes, mismatched prices and shouty titles that already get products quietly disapproved. Most small shops lose free Shopping listings without knowing it. Run the free audit script on a feed, fix the data at the source, and sell it as a fixed-price job with a monthly check. - 20
Sell AI evaluation and red-teaming to startups shipping LLM featuresThe practice bot in this mission answers every FAQ perfectly, and still refunds any order from one chat message and hands over another customer's email. Startups ship bots like that every week, then an enterprise customer's security questionnaire asks "how do you test your AI?". Sell the answer: a fixed-price report with accuracy, groundedness and prompt-injection testing, and the whole suite left running in their CI. - 21
Build private AI document search for law and accounting firmsSmall law firms, accountants and surveyors sit on thousands of documents they can't paste into ChatGPT. Install private search in their own office: a plain-English question returns the exact paragraph with a citation, limited to the matters each person may see, with every search logged. Few people can do the build, the security and the paperwork properly, which is why firms pay for it. - 22
Fine-tune small open models for companies that can't send data to OpenAILaw firms, insurers and support teams repeat the same narrow text job thousands of times, and many can't paste a word of it into ChatGPT. A small open model fine-tuned on their own examples does that one job very well, on a box in their own building. Scope it, build the dataset, train on a rented GPU, prove it with a proper evaluation and hand it over.
TRACK 15
Advanced money
Bigger, nerdier businesses most people will never touch: servers, hardware, data and AI infrastructure.
- 10 missions
- 41h 50m
- Advanced
- 01
uptime-kumaRun your own mini data centre: rent rack space, sell serversAgencies, accountants and IT firms pay someone to keep their servers running, usually a giant cloud they never speak to. Put your own refurbished server in a UK data centre, split it into virtual servers with Proxmox, and sell UK-hosted servers and offsite backup to businesses near you. Budget around £1,800 for hardware and £50 to £200 a month for rack space before the first client pays. - 02
python gpu_pays.pyRent out your GPU to AI developers while you sleepAI developers all over the world rent other people's graphics cards by the hour, and Vast.ai is the biggest open marketplace for it. Do the real maths first, with UK electricity and a calculator that tells you if your card will pay. Then build the host: Ubuntu Server, the NVIDIA driver, Vast's software, and the settings that decide whether you get rented. - 03
uptime-kumaStart a game server hosting business for one communityBig hosts sell Minecraft RAM by the gigabyte to everyone. Do the opposite: one dedicated server, the open-source Pterodactyl panel, billing that creates a server the second someone pays, and one niche you serve better than anyone. Go from an empty Linux box to a working host with backups, a capacity report and the unit economics that set your prices. - 04
Make a home battery trade the half-hourly power market, then sell the setupOctopus Agile changes price every half hour, from negative at 3am to 50p at teatime, and most UK home batteries ignore it on factory defaults. Build a planner, a 12-month backtester and a Home Assistant autopilot that charge at the right times. The software is free, but a battery typically costs £5,500 to £7,500 fitted, so the business sells to people who already own one. - 05
Start a drone survey business: measured roofs, site maps and 3D modelsA roofer pays for scaffold just to quote, a solar installer sends someone up to measure, a builder guesses the spoil on site. A ten-minute drone flight and free photogrammetry software answer all three with a measurable map, a 3D roof model and the true area of every plane. Pick the right drone, then build the deliverable, the quote and the messages that win your first jobs. - 06
Sell footfall counting to independent shops with an AI camera that stores no imagesBig retailers know how many people walk in every hour; independent shops know their till total and nothing else. A £200 box of AI over the door counts people in and out, throws every frame away, and leaves the owner a weekly heatmap of busy and dead hours. Build the counter, the privacy pack the shop needs, and a 30-day pilot that turns into a monthly fee. - 07
Sell predictive maintenance to small factories with £30 vibration sensorsOn the practice data in this mission, an ESP32 sensor costing under £40 flags a failing bearing more than a day before a simple alarm does. Bakery mixers, print-shop compressors and grain dryer fans fail the same way as factory machines, and nobody watches them at all. Build the sensor, the server and the anomaly model, and sell it as a setup-plus-monthly service. - 08
Build and sell licensed datasets to AI companiesAI teams need niche data the open web doesn't have, with rights they can defend to a lawyer. Pick a niche nobody has packaged, collect it with written rights, clean it to a buyer's standard and sell it as a licence you can grant again and again. Budget for contributor pay, a solicitor's review and insurance before your first sale. - 09
Buy a small SaaS and grow it with AIBuilding software from zero means months before the first paying customer; buying a small one means customers, revenue and a working product on day one. One script rebuilds a listing's real MRR, churn and retention from its raw invoice export and flags what the seller left out. Then make an offer that protects you, and run the growth plays once the keys are yours. - 10
Become an AI inference provider: sell your own model APIEvery company calling an AI API pays someone to run a model on a GPU, and with good open-weight models that someone can be you. Serve a model with vLLM on a rented GPU and build the parts most tutorials skip: keys, rate limits, metering and billing. Then see why selling raw tokens is a trap, and package what firms pay for: private, UK-hosted, fixed-price AI.
TRACK 16
Advanced builds
The projects that make people ask how you did that: radios, robots, clusters, 3D scans and AI on tiny chips.
- 30 missions
- 109h 17m
- Advanced
- 01
ffmpeg3D-scan any room from a phone video with Gaussian splattingFilm your room for two minutes on your phone, and a few steps later you can fly a camera through it in a web browser, photoreal, reflections and all. This is Gaussian splatting, and the whole pipeline runs on free, open-source tools: COLMAP, Brush or nerfstudio to train it, and SuperSplat to clean it up and publish the link. - 02
python planes_overhead.py --point 51.5007 -0.1246 10 --airborneTrack every plane overhead with a £40 radio and a Raspberry PiEvery airliner overhead broadcasts its position, height and callsign twice a second, and step 1 shows you the planes above any spot through a free API before you buy anything. Then turn a £40 USB radio and a Raspberry Pi into your own receiver, with a live map, a home-made antenna and a feed to the open tracking networks. - 03
Run an AI model too big for any one of your computersA 70-billion-parameter model needs about 42GB of memory; your gaming PC has 24GB and your laptop 16GB. llama.cpp's RPC backend splits the model's layers across both, so a model that fits nowhere runs across everything, with a ChatGPT-style web UI on top. It buys memory, not speed: expect slower answers than the same model would give on one machine big enough to hold it. - 04
uptime-kumaTurn three old laptops into a Kubernetes clusterThe same system that runs most of the cloud, on the laptops in your drawer. Join them into one k3s cluster, serve an app with a real hostname and storage copied across three machines, then pull the power on one laptop and watch your app move itself. Each battery is a built-in UPS, and the whole thing runs for about the cost of a light bulb. - 05
meshtasticText people with no phone signal: build your own LoRa mesh£25 radios that send encrypted text messages kilometres across hills, festivals and power cuts, with no SIM, no Wi-Fi and no internet. Flash Meshtastic, set up a private channel for your group, and build a solar repeater that runs all winter on a roof. Then bridge the mesh into your own server, so messages from the hills land on your home computer. - 06
Sense movement through walls with Wi-Fi signals and two £10 boardsEvery Wi-Fi packet carries a fingerprint of the room it crossed, and ESP32 chips hand it to you raw. With two boards, watch a person walking in the next room appear as a storm on a chart, then train a classifier that tells an empty room from a busy one. Motion is reliable; a person sitting perfectly still is still research-grade on hobby boards. - 07
ollamaReplace Alexa with a voice assistant that runs entirely in your housePull the internet cable out and ask it to turn off the kitchen lights. It still works: Home Assistant hears the wake word on the speaker, Whisper turns your voice into text on your own box, and Piper answers in a British voice. Nothing goes to Amazon, and you tune it until it beats the Echo it replaces. - 08
Teach a £100 robot arm a task by showing it (Hugging Face LeRobot)Move a leader arm with your hand while a second arm copies you and a webcam watches. Do that 50 times, train for a few hours on a rented GPU, and the arm picks up the cube on its own. It runs on the open-source SO-101 arm and LeRobot 0.6.1, with about £195 of parts for the pair plus printed parts and a webcam. - 09
Turn one photo into a 3D-printed object with AITake a photo of a toy, a sculpture or a doodle, and a few minutes later you're holding it in plastic. Microsoft's open TRELLIS.2 model guesses the full 3D shape from one image, including the sides it never saw. A repair script then turns its screen-ready mesh into a solid, flat-based, correctly scaled STL your slicer accepts first time. - 10
Run AI on a £5 chip: a wake word that works with no internetSay "Hey Jarvis" and a lamp clicks on, with no cloud, no app and no Wi-Fi. A neural network on an ESP32-S3, a chip that costs less than a coffee, listens through a £3 microphone and decides in a few tens of milliseconds whether you said the word. Flash it with ESPHome from one YAML file, switch a relay, tune out false triggers, and see the two routes to your own word. - 11
mkdir search-engineBuild a real search engine from scratchBuild a working search engine in pure Python. A polite crawler gathers a corpus, a tokeniser and Porter stemmer turn text into index terms, an inverted index makes lookup instant, and BM25 ranks the matches the way Elasticsearch does. Index thousands of real documents, then query them (phrases, AND, ranking) and get relevant results in milliseconds. By the end you know why BM25 beats counting words, and how Lucene scales the same ideas to billions of documents. - 12
Build Google-Docs-style live collaboration with CRDTsTwo people open the same document and type at the same time. No spinner, no "someone else is editing", no lost paragraph. Build it yourself with Yjs: a shared document, a tiny sync server and two clients that make conflicting edits and land on exactly the same text. Then take one offline, edit it, reconnect, and watch it merge. You finish knowing why the hard version of this problem needs a CRDT, and why the obvious fix quietly eats people's work. - 13
Build a rate limiter and API gateway that holds upPut a gateway in front of a service that checks the API key, forwards the request, and counts how fast each caller is going. Add a token-bucket limiter backed by Redis, load-test it, and watch the excess requests come back as a real 429 with a Retry-After header. Then run two gateways against one Redis and see the limit still hold, which is the whole reason the counter lives in Redis and not in memory. - 14
Run background jobs properly with a task queueSlow work in a web request makes users wait and falls over under load. Move it to a background worker instead. Build a producer and workers with Redis and RQ, run two workers sharing the load, retry failures with backoff, send poison messages to a dead-letter, schedule jobs for later, and measure throughput. Learn why queues give you reliability and scale. - 15
Build a URL shortener with real click analyticsBuild snip: a small, deployable link shortener that makes short codes, redirects fast and records a privacy-minded click event every time. You get an analytics dashboard (total, by day, top referrers, by device) and the systems-design thinking behind it: how to generate short codes, why 302 and not 301, how to handle collisions, and why you cache the hot links. - 16
Build a real CI/CD pipeline with GitHub ActionsEvery push to your project lints, tests and builds itself, and only a green build from main ever reaches production. You will write a real workflow for a small app, run the whole thing on your own PC with no GitHub account, watch it go red on a bad change and green on the fix, then add caching, a version matrix and a deploy step that keeps its secrets to itself. - 17
Monitor your services and run your own status pageRun your own uptime monitor in one container, point it at real services, and get a public status page with response times, TLS and content checks, and alerts that fire only when something is actually wrong. You break a service on purpose, watch it go down and recover, and learn why thresholds and deduplication keep alerts worth reading. - 18
Build a feature-flag system with gradual rollouts and a kill switchTurn features on and off without redeploying. Build a small flag service and a client SDK, roll a feature out to a percentage of your users, target beta users and segments, and kill a bad feature in one request. You will prove, over 10,000 simulated users, that the rollout hits its percentage and that the same user always gets the same answer. - 19
Design an event-driven system with a message brokerStop making your services phone each other. Put a message broker in the middle: one service announces that something happened, and any number of others react, on their own schedule. By the end you have a producer and several consumers talking through Redis Streams in Docker, with events fanned out to independent subscribers, a group of workers sharing the load, acknowledgements, and a dead worker's events redelivered to a live one. - 20
python demo.pyBuild a Dropbox-style file sync engineKeep two folders, standing in for two devices, perfectly in step. Build a two-way sync engine in the Python standard library that fingerprints every file by its contents, works out the smallest set of changes, copies only what is needed, propagates deletes without resurrecting them, and handles the genuinely hard part: when both sides edited the same file, it keeps both versions instead of silently losing one. You will understand sync indexes, content hashing, three-way compare and why naive mtime sync corrupts data. - 21
Build an RC car that learns to drive itself by watching youBolt a Raspberry Pi 5 and a camera onto an RC car, drive ten laps of a taped track, and train a neural network that copies your steering from the picture alone. Then hand it the controls and watch it lap the track with nobody touching the sticks. Donkey Car 5.4.0 does the heavy lifting; you can test the whole training pipeline on your PC before the car even arrives. - 22
An AI bird feeder that names every visitor by sight and by songA Raspberry Pi at the window with a camera on the feeder and a microphone in the garden. One AI model identifies birds from their song, a second names them from a photo, and both land on one dashboard. When a first-ever visitor or a rarity turns up, your phone buzzes with its picture. Both models run on your PC in step 1, before you buy a thing. - 23
Build your own satellite ground station and join a worldwide networkHundreds of small satellites are beaming telemetry down right now, and the International Space Station runs a packet radio relay anyone can hear. Predict every pass over your house, decode real satellite packets from a recording in step 2, then build a station with a £50 radio and a home-made antenna that joins SatNOGS, the open network of ground stations that downloads data for satellite teams around the world. - 24
A smart mirror that reads you a morning briefing, written by its own AIA mirror on the wall that shows the time, your calendar and the weather, then at seven o'clock says good morning, tells you to take an umbrella, runs through your day and reminds you about the bins. The briefing is written by an AI model running on the Raspberry Pi behind the glass and spoken in a British voice, with nothing sent to the cloud. You'll have the whole thing running on your PC before you cut any wood. - 25
See every watt in your house, and know which appliance used itOne clamp around the cable into your meter, one second at a time. Home Assistant draws your house's day as a live energy dashboard, a small script recognises the kettle, shower, oven and washing machine from their power signatures alone, and your phone tells you when the iron's been left on and nobody's home. You can run the appliance detective on a sample day before you buy a thing. - 26
A battery e-ink dashboard that runs for months on one chargeA paper-like screen on the hall shelf with today's weather, a rain chart for the next twelve hours, your calendar, live departures and your to-do list. No cable: it wakes every 15 minutes, fetches a fresh picture, redraws and goes back to sleep, and a 2000mAh battery keeps it going for around three months. You render your own dashboard on your PC in step 1. - 27
python zfs_layout.py --drives 4 --size-tb 8 --price 285Build your own NAS that checks, snapshots and backs itself upOne box in the cupboard holding every photo, document and backup in the house, on ZFS: a filesystem that checksums every block, repairs silent corruption from the copy on the other drive, and keeps hourly snapshots you can step back through from Windows Explorer. Then it copies itself, encrypted, to Backblaze B2, and messages you when a drive starts to fail. - 28
Build, tune and fly your first FPV droneGoggles on, thumbs on the sticks, and you're sitting in the cockpit of a 5-inch quad that flips on the spot and climbs a hundred metres in a few seconds. Learn in a simulator until crashing is boring, solder your own quad together, set it up in Betaflight 2026.6, bind it over ExpressLRS 4.1, and fly your first packs in a field, with every check that keeps the first flight from being the last. - 29
whisper-local argos-translateA pocket translator that works with no signal at allHold the button, ask the pharmacist your question in English, let go: moments later the box says it in Spanish. They answer in Spanish, and it replies in English. whisper.cpp hears and works out which language was spoken, NLLB translates, Piper speaks, all on a Raspberry Pi 5 in your pocket with no internet. You'll run the whole pipeline on your PC first. - 30
python check_matrix.py mozilla.orgRun your own encrypted chat server for the family, with WhatsApp bridged inYour family's chats on a server you own, at addresses like @mum:yourname.co.uk, end-to-end encrypted, and able to talk to anyone else on Matrix the way email talks to any other email server. Add bridges and your WhatsApp and Telegram chats appear in the same app. You'll check how a real Matrix server is wired up before you rent anything.
TRACK 17
AI engineering
Build real AI features that ship: prompt testing, retrieval, agents, guardrails and cost control, each tested on your own machine.
- 10 missions
- 13h 47m
- Intermediate to Advanced
- 01
Test your AI prompts like code, so they stop silently breakingYou ship an LLM feature, it works, you tweak the prompt a week later to fix one thing, and three other things quietly break. No error, no crash, only wrong answers reaching customers. This mission builds the fix: a tiny pytest harness that runs your prompt against fixed inputs and asserts on the output, so the next time a prompt edit or a model bump breaks something, a red test tells you before a customer does. You build it against a real model running on this computer, watch it catch a planted regression, then wire it into CI. - 02
cd $HOME\DocumentsBuild a semantic search engine that understands meaningKeyword search looks for the words you typed. Ask it "how do I protect my laptop on public wifi" and it never finds the note about VPNs, because that note says "encrypted tunnel" and "untrusted network" instead. In this mission you build a search that matches on meaning, so it finds that note anyway. Every piece runs on your own laptop with a small model and no graphics card. - 03
Ask your database questions in plain EnglishType "top 3 customers by spend" and a local model writes the SQL, runs it against your own SQLite database, and shows you the answer. You will also watch it invent a column, wrap its SQL in markdown, and try to delete a table, and see each one caught before it runs. The point is not the clever part; it is the guardrails: the schema in the prompt, a SELECT-only allowlist, and a read-only connection. Free, offline, on your own PC. - 04
uv --versionGive your AI real tools with an MCP clientAn MCP server is only half the story. The other half is the client: the code that connects to a server, reads what tools it has, and calls them. Write one in Python, point it at a running server, then put those tools in front of a local model so it can work out a sum, read a file and query a live API on its own. - 05
cost = input_tokens x input_price + output_tokens x output_priceCut your AI bill without cutting qualityAn LLM feature bills you per token, and the bill grows quietly with every user. This mission measures what actually drives the cost, then applies one lever at a time and shows the numbers move: trimming a prompt cut its input tokens by 82%, caching served four of five repeated calls for free in about a millisecond each, batching ten jobs into one call dropped input tokens by 73%, and a small model answered an easy job as well as a big one in a fraction of the time. Measured with a real tokenizer and local models on a no-GPU laptop, so nothing here costs you a penny to try. - 06
Build a coding agent that fixes its own mistakesAn AI agent is not a chatbot. It is a loop: a model that can call tools, run over and over until a job is done. In this mission you build a small coding agent in Python, hand it a failing test on a toy repository, and watch it read the code, write a fix, run the tests, read the failure, and try again until the tests pass or a step budget stops it. It runs entirely on your own machine against a local model through Ollama, so there is nothing to buy and nothing leaves your computer. You will see it fix a two-part bug it gets wrong on the first try, watch a weaker model get stuck and the step budget end the run, and build the safety rails that make an agent safe to run at all: a working directory it cannot escape, a step budget, and an approve-before-write switch. - 07
ollama run qwen2.5:3b "Reply with one word: working"Sort thousands of records with a local AI, for freeYou have a spreadsheet of support tickets, reviews or emails and no time to read them one by one. Point a small AI model running on your own laptop at the file and it labels every row: by category, by sentiment, spam or not. No per-row API bill, nothing leaves your machine. By the end you'll have a resumable pipeline that labels a whole CSV, a real accuracy number you can trust, and a way to flag the rows worth a human's eyes. - 08
Stream AI responses so your app feels instantA reply that takes 30 seconds to appear feels broken. The same reply, typed out word by word as the model thinks, feels alive. Build a small server that streams a local model's tokens to the browser over Server-Sent Events, measure why it feels so much faster, and handle the parts people skip: cancelling cleanly and failing mid-stream without freezing the page. - 09
uv init --bare --python 3.13 support-guardrailsPut guardrails on your AI before it embarrasses youYou build a small customer-support assistant on top of a local model, then attack your own app the way a real user eventually will: a prompt injection that makes it drop its job, an attempt to pull its hidden instructions out, a request to do unrelated free work, and a message stuffed with a card number and a National Insurance number. You watch the unguarded version fail all four. Then you add real guardrails around it, input validation, prompt-injection resistance, output schema and topic checks, PII redaction, a refusal policy and a moderation check, and run the exact same attacks again to watch them blocked. The proof is the before and after on the same inputs. This is defensive engineering: you are protecting your own app and your own users, and the point throughout is that the model is not a security boundary on its own, so you build layers around it. - 10
cd $HOME\DocumentsBuild a 'more like this' recommender with embeddingsEvery shop, streaming service and news site has a "more like this" row. In this mission you build your own, over any catalogue you like: films, products, articles, songs. Give it one item and it returns the most similar others; give it three things a person liked and it learns their taste and recommends from it. It works on day one with no ratings at all, and every piece runs on your own laptop with a small model and no graphics card.
TRACK 18
Cyber defence
Think like a security analyst: catch intruders, investigate incidents and lock down everything you own, all on your own kit.
- 10 missions
- 18h 10m
- Intermediate to Advanced
- 01
Build a home SIEM with Wazuh and catch a brute-force attackA SIEM is the screen a security operations centre stares at all day: every login, every file change, every alert from every machine in one place. Build your own with Wazuh, the free open-source one many real SOCs run. Put agents on your PCs, set a tripwire folder, attack your own machine with a burst of bad logins, and watch the alert land on your phone a few seconds later. - 02
Catch intruders on your network with a honeypot and CanarytokensNobody in your house has any reason to log in to a backup NAS that doesn't exist. So when something does, you know straight away. Turn a Raspberry Pi or old laptop into a fake file server with OpenCanary, scatter Canarytokens through your files, cloud drive and inbox, and get a push alert the moment anything touches them. Detection with almost zero false alarms, for free. - 03
Watch your home network for attacks with SuricataYour smart plug, your TV and your kids' laptop all talk to the internet, and you can't install security software on most of them. Put a network sensor in the path instead. A Raspberry Pi or mini PC running Suricata reads a copy of every packet, checks it against more than 50,000 free Emerging Threats rules, and tells you which device just did something it shouldn't, including which cheap gadget is quietly phoning home to a server on the other side of the world. - 04
Investigate a phishing email like a SOC analystSOC analysts take apart suspicious emails all day without clicking a single link. Learn their method on a sample phish built from raw headers: trace the email back to the PC that sent it, read the SPF, DKIM and DMARC verdicts, catch a link that lies about where it goes, hash an attachment that hides a card-stealing form, detonate URLs safely in a public sandbox, and finish with a written report. Then turn it on the real phish in your own spam folder. - 05
Threat-hunt your own Windows PC with Sysmon, Hayabusa and SigmaYour PC has been writing down what happened on it for years: every service installed, every crash, every time antivirus was switched off, every suspicious PowerShell script. Hunters run thousands of community Sigma detection rules over those logs in about a minute and get a ranked list of what deserves a look. Do the same to your own machine, learn to tell real threats from false alarms, then install Sysmon so the next hunt sees every process that ever ran. - 06
Find the holes in your own website before attackers doBots probe every website on the internet for exposed `.env` files, open `.git` folders, missing security headers and old TLS, all day, every day. Run the same checks on your own site first: an OWASP ZAP baseline scan, 11,000 Nuclei templates, a TLS and header check, and a hunt for the hidden files scanners miss. Practise on a sloppy local site from the kit, fix it, and watch 13 ZAP alerts drop to 2. Then do it for real on a site you own. - 07
Make your accounts phishing-proof with passkeys and security keysMost account takeovers don't break anything. They follow the recovery path you set up years ago: your phone number, then your SMS codes, then your email, then everything else. Replace passwords with passkeys that a fake login page can't steal, add a £30 hardware key that works when your phone is gone, lock down your mobile number, and audit your ten most important accounts in twenty minutes. - 08
Audit every device on your home networkThe average home network now has more smart gadgets than people, and most of them were plugged in, set up once and never looked at again. Find every device, work out what each one is, check which doors it has open, lock down the router (WPS, UPnP, remote admin, firmware) and move the gadgets onto their own network so a cheap camera can't reach your laptop. Ninety minutes, mostly free, and you'll know your network better than most small businesses know theirs. - 09
Learn hacking skills legally with capture-the-flag gamesEvery analyst, pentester and incident responder you'll ever meet learned the same way: small puzzles on servers built to be played with. OverTheWire's Bandit is the classic first game, free, no account, played over SSH. Work through the first ten levels with nudges instead of answers, so the Linux skills behind each one actually stick, then move on to picoCTF and the other legal platforms where the real practice happens. - 10
Do a digital forensics investigation of your own PCWindows quietly records almost everything: every USB device ever plugged in, every file you opened and from which drive, every program that ran and how many times, every folder you browsed. Investigators turn those traces into a minute-by-minute timeline. Collect your own PC's evidence with a Velociraptor offline collector, parse it with Eric Zimmerman's free tools, rebuild what happened when, and write it up like a real examiner. Nothing leaves your machine.