Ideas (June 2022 edition)

· 2299 words · 11 minute read

Various geek ideas, maybe startup ideas. Some are ones I’m thinking about, and some are ones I wish you would think about.

  • Solve the Sybil Attack problem. An online game bans a player for being a jerk, but the person creates new “sockpuppet” accounts and keeps being a jerk. It’s a hard problem to solve because it pits decentralization, privacy, and scalability against each other. Most solutions I’ve seen favor scalability over decentralization and privacy. For example, Target won’t sell me certain cough medicines unless I let them swipe my government-issued driver’s license. This tradeoff isn’t surprising; people typically don’t care about privacy or decentralization until they’re gone, and business and government entities derive a lot of value from knowing more about their customers. This is also why a privacy-respecting solution will have a hard time getting adopted. Why would a store prefer to merely know that I haven’t purchased too much pseudoephedrine in the past month, when they could instead record my name, home address, birthdate, and a unique identifier that lets them track my purchasing behavior for the rest of my life? For a better-balanced solution to be successful, it will have to create incentives for everyone to participate, and for individuals to demand its use over the lazy “can I see your ID?” solution. I’d like to see such an incentive system.

  • Build a groovebox. I have an obsession with music production. Or, rather, getting into music production. I’ve tried and failed probably 20 times over my life to get past the learning phase and into actual composition and production. I’ve taken piano and guitar lessons, I’ve completed online courses about music theory, I played keyboards in a high-school band, and I’ve amassed way too much music-production software and hardware. Any reasonable person would give up on this dream, but I’m apparently not reasonable. About a year ago I started thinking about building a groovebox. I realized I have relevant skills (software, basic hardware, familiarity with music theory), and that a groovebox is a little like a mechanical keyboard, several of which I’ve built from scratch (as in sending PCB designs out for manufacture). I envisioned something in between Pocket Operators and an Akai MPC, but more powerful than the Novation Circuit. Then I found the Dirtywave M8, which was very close to what I had in mind. And that product was designed and implemented by just one (very talented) person! I think there’s room for more than one such product in the space.

  • Private public restrooms. I remember something like this in Snow Crash. A public restroom that somehow restores itself to a pristine state after you use it. You pay to use it via an app, maybe by subscription. You agree to stringent terms of service beforehand that you won’t do anything gross in it. Before you exit, you sign off on automatic photos of the empty room that show that you didn’t vandalize it. Challenges: someone goes in there and sleeps. Someone tailgates after you and then does bad things. It malfunctions and automatically opens while you’re indecent. Someone hacks the security cameras and publishes pictures of you pooing. It ends up being not clean enough to beat Starbucks. The community decides that this calls for class warfare. Related: Japanese capsule hotels, which are probably dependent on elements of local culture to be successful.

  • Gamified public restrooms. On the outside of the door is a timer, counting up, that resets every time the door closes. Maybe there’s a daily leaderboard for fastest and slowest trips. Inside is another timer display and maybe a beep that sounds when you advance up the slow leaderboard. Using a camera or sensors on the floor, also let the person inside know how many people are waiting. Ooh, maybe you can join a clan!

  • Code up the tax code. Build a portable engine that represents the US (state and federal) tax codes. Insist on great unit tests. Encourage people to run it, turn it into apps and services, whatever. Do to Intuit what Stack Overflow did to expertsexchange.com. Eventually the expectation becomes that the more objective parts of legal code are expressed in consistent and machine-readable fashion, hopefully leading to more consistency, less repetition and duplication, and a cleaner separation between the toil of compliance with bright-line rules and the genuinely difficult remainder.

  • Hardware secret protector. More self-contained than a Yubikey, more general-purpose than a Ledger Nano, more usable than a Mooltipass, but a lot less complex than a smartphone. It would fit on a keychain. I could use it to pay for things by NFC, to withdraw money from an ATM, and to authenticate to websites. Bringing multiple of them together enables Shamir’s Secret Sharing magic. Losing it or breaking it would be an inconvenience, but not disastrous; there would be a way to generate a new one from an existing one or from a master secret. Probably uses biometric authentication, and maybe a short-lived token that erases when the device is separated from the owner or the owner uses a duress gesture.

  • Append-only note-taking app. A combination of a personal assistant and a private Twitter. Throughout the day I add short snippets of text. The system incorporates them into a dashboard of todos, meeting notes, document outlines, etc. It’s connected to my calendar and address book, so it knows that a note entered at 11:14 am is related to the meeting I’m in at that time. The append-only aspect is important because it implies that I can send input from anywhere, without looking at the UI, so I can scribble a note during a call or a lunch meeting without breaking the flow.

  • ML therapist. A synthetic chat buddy. It remembers basic facts about you and can carry on a decent conversation. Utter a magic incantation, and it deletes itself. Bonus points if it lives inside a physical device that you can hold in your hand, so you know where your secrets are.

  • Paired pairs of earbuds. Two pairs of earbuds. If I speak into one, the other pair hears my speech. They otherwise act like regular earbuds. A couple wears them while they’re out in public. They can have a conversation with each other without constantly pausing whatever they’re listening to. Could work really well for a family sitting within Bluetooth range on an airplane.

  • Turn your home printer into a drag-and-drop PDF-to-paper converter. A Raspberry Pi image that you install, configure once, and plug into your printer’s USB port. From then on, printing is a matter of sending PDFs to it. No driver hassles or OS compatibility issues. I wish Google Cloud Print had gone this direction rather than dying completely.

  • Escape rooms on phones and VR. This must exist already.

  • Relaxing VR spaces. I love the Quest’s waiting rooms. I would buy more if they were for sale. All I’d want to be able to do is wander around and look at things. If a startup could develop a style guide that artists could learn, then it could monetize cheap art labor from around the world. Maybe invite your friends to see your crib. Obligatory #nft keyword.

  • Inventory app. Google Photos does an OK job of recognizing things in photos. Search for “dog” or “flower” and you’ll get your pictures of those things. I’d like to be able to take a picture of the knick-knack drawer in my kitchen, and have ML recognize everything in it. Over time that picture would become stale, but I could refresh it by taking a new picture. Same with the refrigerator, and the medicine cabinet, and all the shelves in the garage. Once I’d taken pictures of everything in the house, I could ask “where is the screwdriver?” and it would reply “I last saw it on the left side of your workbench.” Perhaps the app is connected to Amazon, Wal-Mart, Target, and other places I shop, so that it knows my shopping history and has a better idea what the pictures are likely to contain.

  • Board-game scoring app. Take a picture of a board game in progress. The app tallies up the scores and maybe even gives suggestions for the best next move.

  • Board-game hardware platform. Often attempted, but AFAIK there is no successful board-game-console platform in existence. A physical board that can become any board game via DLC (downloadable content). Feedback from my board-game-fan friends is that the market’s too small, the game pieces are too varied, the software threatens to take away some of the magic, and game developers have little incentive to participate in a rent-seeking platform. These are all good points.

  • Dashboard appliance that doesn’t suck. Install a .deb on a fresh Raspberry Pi device and reboot. The boot screen shows a QR code. Visit that URL on another computer and start customizing the dashboard. That’s it. I haven’t found anything that takes care of autostart, fullscreen, and all the other configuration that an appliance use-case should handle. There are enterprise dashboard services out there, but they’re priced accordingly, which isn’t worth it for me to see local weather and Bitcoin prices.

  • Another little computing ecosystem. Pico-8 and uxn are examples. I don’t know whether the world needs another one of these, but it would be a lot of fun to develop one. Maybe combine this with the groovebox idea so people can share their VST-like plugins and effects.

  • Smart dead-man’s switch. If you don’t ping this URL within X time, then it executes a script you defined. How does a breach of the database not result in a horrible disaster for lots of people? Same idea but only once the URL is pinged. Database compromise is an easier problem in the second case, as the URL-ping could include a secret key.

  • Key/value store as a public service. The URL contains a public key. Anyone with that URL can read it. Only someone with the private key can formulate a POST that writes to it. Stale URLs that aren’t used within X time are purged. All sorts of spam and illegal-content problems. I don’t know who would use this, but I do think it’s interesting to try an auth system based on public-key cryptography.

  • Shared todo-list templates. If I add “go on trip” as a todo, it offers to expand to 20 sub-items that most people would need. Pack toothbrush, find passport, pack charger for laptop computer, etc. Could appeal to Checklist Manifesto fanatics. The list template becomes a unit of content, like YouTube how-to videos.

  • A trustworthy app brand. Utility apps that are consistent, simple, and free. Compete with all the ad-supported garbage out there.

  • Modular IoT ecosystem. Setup is open and well-specified. Devices self-document their RESTful communications schemas. Home Assistant and ESPHome are getting there, but very few device manufacturers build exclusively for that ecosystem; rather, they tend to require their own awful mobile apps for setup, and a good samaritan in the community reverse-engineers the protocols and creates an HA adapter. I’d like things to tip to where open is the default, and the proprietary plays need to demonstrate why they should exist at all.

  • IoT intercom. Push-to-talk speakers for a family. Admit it, sometimes you text-message people in your house, and when you do, you feel like you’ve failed as a human being. Using your voice to communicate isn’t any worse; in fact, it might be a little better.

  • IoT space heater. Default off. Pinging a certain endpoint turns it on for 15 minutes. If you pinged it every 14 minutes, then, it would stay on constantly. The point is to make it connected, but simply and safely.

  • Self-recharging crash-proof indoor drone. It knows your house layout and follows a defined path. It has enough sensors to avoid bumping into people; in fact, maybe it flies inches above the ceiling to reduce the risk of collisions. It lands on a charging pad, so you never have to touch it. Why would you want this? It could be a (loud) security device. It could tell you whether you left your purse on the bed or forgot to turn off the oven. It could probably even water the plants, one tiny spritz at a time.

  • Identifiable luggage. It has a built-in RGB LED strip, a BLE radio, and an accelerometer. When it is jostled, the BLE radio starts listening. If someone asks it to via a GATT service, then the LED strip blinks a rainbow for a few seconds. This helps you identify your luggage at the airport baggage claim. The knee-jerk criticism is that TSA will decide this is a bomb. I hope that in the 15 years since that stupid incident, relevant community standards have improved.

  • Wholesale or white-label storage units. I provide the real estate, the structure, the locking system, an API, and maybe an app. You lease the premises from me and market it however you want. It’s a way to amass monetizable commercial real estate.

  • Approval as a service. When this API gets pinged with metadata, a human looks at the metadata and then decides yes or no. I presume this already exists for ID verification and similar, but not as a general-purpose service. Over time you train ML to answer for you, either as a second opinion to the human’s opinion, or taking humans out of the loop entirely. How do you keep the humans from surreptitiously saving private data? Is this just Amazon Mechanical Turk, but more twitchy?

  • Sign-making app. Real simple: I type a phrase, and it shows it it big, readable letters in fullscreen on my phone. Useful when I’m trying to communicate with a family member surrounded by sleeping people on an airplane, or someone talking on the phone. In the old days I’d scribble on a piece of paper.