Ask HN: Impossible Ideas?
Hello HN,
Is there any idea that you have that is great, but feel that it is impossible at the present.Or may be an idea that you think that the incumbent or competition is super strong.
I am just looking an idea to hack on weekends after being bored of creating similar apps through out my IT career. Yeah, why does clothing still come in S/M/L? Why can't I order existing brands in styles 100% fitted to my unique body. Surely you can instantly measure shoulder/chest/waist width, torso/leg height, etc with technology like Xbox Kinect. And on-demand, custom clothing has to be solvable in 2020, right? So it's just a momentum/supply chain problem, like Netflix vs old guard? https://www.eshakti.com/ does this for some clothes (not via Kinect, you have to enter the measurements). No personal experience, but it's nice to see folks trying it. It's baffling to me that Amazon doesn't have something like this. Every vendor of clothing seems to upload a size chart or puts it in the text description, and you wind up with stuff like "6x plus size dress" having a waist size of 32 inches. It looks like someone is thinking about this at Amazon: https://qz.com/963381/amazon-amzn-has-patented-an-automated-... I think that's probably a ways off. I don't understand why someone can't put in a few basic measurements - waist, hip, bust, inseam - and filter clothing based on those, today. It wouldn't be perfect, but it'd be a damned sight better than putting "plus size dress" and getting keyword-stuffed listings made for a size 0. IMO, it's a unit economics problem. Mass producing/selling S/M/L is much more profitable than selling extra wide shoulders/small ribcage to two people. Doing this would require on-demand manufacturing with operational costs close to what currently exists. I can see how it could end up cheaper simply due to not making clothes that don't get sold. I'm sure there's a lot of waste currently. Already here: https://www.sonofatailor.com/ Not a F but I understand lots of them would be very interested in bras that fit. yes, that already exists.... I used MTaylor, a while ago (download the app, scan your body via the camera), and the shirts came in perfectly fitted to my body. The jeans, not so much (they still need to work on it), but for shirts they were spot on. I will never buy 'off the rack' shirts anymore. I'm familiar, but MTailor is a standalone brand. I'm dreaming of a world where I can see any piece/brand of clothing and not even think about the size or fit. It's just taken as a given that I can see it on myself (AR?) and get one that fits me perfectly. The clothing outlets, last stalwarts of strip malls, would be a thing of the past. No such thing as needing to "try it on" anymore. Everyone has different fitting issues. My main one is waist length/width and shoulder width. I'm 6'2" and fit so the waist of most shirts barely go below the top of my jeans and getting the shoulders to fit often means having a parachute around my waist. It looks dumb and clothing fit is extremely important to overall appearance. There's a lot of disruption to be had in this space. Edit: to be slightly more clear, I'm thinking of technology to replacing the incumbent infrastructure/processes/consumer experience, as opposed to just creating a new specialty brand. Does it only work in the US or is there an offering for EU/Germany as well? Ordering pants online is a nightmare, as retailers do not know what length and width are. wasn't there a startup on shark tank that did this, rejected the sharks' offers, then bootstrapped to success? I think it was MTaylor. My current favorite shirts came from them. It's MTailor (like a tailor) and I think they're great, but my dream is a bit different (elaborated in another comment). I’ve long been interested in the “cocktail party problem” which involves disambiguating audio in a conversation with multiple people. I think this tech is foundational for better video calls and smart speaker devices for homes. The best research I’ve seen on this is from Mitsubishi but as far as I know this is well into the territory of an impossible problem today. There's been a ton of interesting research and progress on this in the last few years: https://ai.googleblog.com/2018/04/looking-to-listen-audio-vi... An RSS reader that can organize feeds in a tree view. So if HN and Bruce Schiner's blog both link to the same article, that article would be the parent node and they would be two items off of it. If there were responses and followups the tree could grow larger. Even bigger: Discovering missing parts of the tree. Creating a desktop OS that feels free to break past UNIX standards and start from a clean slate Or at least a usable alternative to Microsoft, preferably from a different country. There isn't really competition on that market. At least in Europe I don't feel like macOS makes even a blip in the business market. Temple OS seems impossible, and yet exists. Urbit is a pretty good example of this. Fuchsia? I believe Fuchsia is targeted towards ARM chips(mobile devices)? Plan 9? Open source Pokemon, where new creatures can be contributed by anyone and anyone can add stories and adventures or fork the whole thing and do whatever they want. The main reason for the commercial success of Pokemon is the huge array of readily available merchandise and anime/movie tie-ins. How would you go about replicating that with a crowd-sourced competitor? How would I be able to obtain a plushie for my favorite crowd-sourced character without paying the high prices associated with custom commissions? It'd be open source so you'd get plushies, figurines, videos etc after someone else made them because permissive licensing let them use the art and ideas to do whatever they want. There would be nothing preventing a plushie company or a person or a DIY project making them affordable. Yeah, but economies of scale couldn't apply as much, increasing unit prices. But competition could be much higher, pushing down unit prices... It would be potentially much cheaper without licensing negotiations, royalties and IP policing adding to the price too. It would need to be curated, though, the way levels in Mario Maker are. Otherwise you'd just end up with people posting porn and ads or troll creatures designed to ruin everyone's fun. I want a second, locked-down internet/web that coexists with the current internet. The type of controls: - A decentralized group of "meatspace" businesses provide notarization. To get an "account" to use the network and its websites, it is mandatory that a person provide government ID. Users can allow a notary to make arbitrary types of information about themselves public (eg: prove they once had cancer,prove they travelled to Cuba, prove they have a high-school diploma, etc). "Private" info is still available to law-enforcement. This makes users more accountable, both for breaking laws, and for (to use the phrase loosely) stolen valour. It might be useful, in addition, if users needed to provide some deposit as collateral in case they abuse notarization. - The web piece of the network uses a protocol with no page- or site-provided style sheets. A developer can create a style-sheet, but the user chooses styles per-browser (eg: the same style-sheet for all websites they view). This makes for a dull, but usable, web-browsing experience. No font-sizes changing, articles laid out the same everywhere, widgets work the same everywhere. - The web piece of the network uses a protocol with no page- or site-provided dynamic features (ie: no scripting). Any dynamic features (eg: a login box, or an upload status bar, etc etc etc) need to be standards. The group responsible for this new web protocol can always treat the old web as a playground from which to steal ideas. This makes shady advertising, and many types of hacking, impossible. Code editor that fully embraces power of VR, and treats code not as text, but as AST. Imagine every node of the AST on it's own plane, with every related signature or declaration just one vim-style keystoke away, opening right next to the previous one, instead of above it. There's nothing impossible from the technical side, but figuring out what is actually useful and time-saving in terms of UX and what would just end up as useless eye candy would require enormous amount of work, I'm afraid. Decentralized, unique, human-usable identifiers, free and anonymous to legitimate humans (limited to only a few per) but unobtainable by non-humans. I can accept humans giving/using up some of theirs for non-human purposes. Not really an idea, but seems impossible. $1 scratchers with a guid on them at every corner store might have some use. I want someone to build a better linkedin. A database where I can search "people I know that speak French." Or "people I know that know something about digital marketing". Or "people I know that know how to cook well."
Everytime I have gone to linkedin to try and find someone that has a particular skill, it has failed me. This is probably challenging to do well because linkedin is already trying to do this, and because of privacy issues. Not an impossible idea, but project management for life. Construction project at home? Punchlist, budget, progress tracking, vendor relationship, contract management, dependencies and blocking. Comparison shopping? Gift lists, family research, nominations, voting, chipping in for purchases. Cooking at home? Family nominates a recipe for dinner, cross reference pantry, add needed items to shared grocery list. Transportation? Whos taking what kids to what events, in the event a plan changes, a push notification goes out to all parties (friends, drivers not in the family.) Budeting? Projected cashflows. Asset allocation in multiple accounts is incredibly complex, with different family members having different roles. Being able to budget and save, while still giving extended family privacy. Signing a kid up for soccer, or starting a new construction project automatically updates the budget. Calendars exist, task management like asana exists, financial management like mint and personalcapital exists. But tools are either too generalized or too specific to interact with other ultra specific domains. An app knowing when someone leaves work, knowing they are the best person to pass a grocery store and pick up kids en route from soccer. We have all sorts of complex interactions in life that require planning, and they are often extremely repetitive sequences. The apps that do exist in these specific fiends often dont facilitate collaboration with other families. tldr: instead of generic task management, management with detailed workflows, workflow interactions, automation, consensus building. You read my mind, ilolu. Lately, I've had so much time in my life that I've been literally coding out of boredom. I am a great engineer with 15+ years of experience doing hundreds of different things. I do not need money. I need to feel, once again, that I have made the impossible, possible. If you wish to get in touch send me an email to hn @ <username>.com. I leave it here from time to time and a few folks have decided to write back. One day, hopefully, we could be a community of sorts. A decentralised pseudonymous reputation system, so that your good-standing in one online community (platform/silo) can be visible to people in another online community. The answer might involve some sort of web-of-trust solution, and storing proofs on a blockchain, but the hard problems are how to avoid Sybil Attacks and not exposing people's social graphs. Isn't it one of the goals of Urbit? I guess, but: "Urbit OS is a completely new, carefully architected software stack: a VM, programming language, and kernel designed to run software for an individual." Even solving an impossible problem shouldn't require all of that. Sure, but nobody knows how to use Urbit. For a few years I've thought it would be cool to experiment with neural interfaces for bomb sniffing dogs/cadaver dogs, etc - so instead of training the dog to find several specific compounds, the system could be trained for multiple compounds. You wouldn't be training the dog, but the system instead. Impossible to do with my current resources. Seems like something that will probably happen eventually. Some scammer made ~80 million bucks selling fake bomb detectors https://www.bbc.com/news/uk-22380368 - I wonder if someone actually made the real deal if it would even be as profitable. > Some scammer made ~80 million bucks selling fake bomb detectors Sadly we see the same devices being used to "detect" covid-19. https://www.independent.co.uk/news/world/middle-east/coronav... Flying cars. It's 2020. Where are the flying cars? I don't mean giant drones either. I'm talking honest to goodness anti-gravity floating vehicles. I love watching the stories of scientific discovery on the reboot of the Cosmos series (https://en.wikipedia.org/wiki/Cosmos:_A_Spacetime_Odyssey) and I can't help but feel we lack a true understanding of gravity. I'm waiting for that one brilliant thinker to have a eureka moment with a simple experiment that makes it all very clear. > I can't help but feel we lack a true understanding of gravity. We have a very solid understanding of gravity [1]. It's just that it's not as accessible as you and I would like it to be [2]. Unfortunately, you cannot argue aesthetics with the laws of nature. [1] Source: When majoring in physics, one of my exams was on general relativity theory. [2] As in: Even solving simple symmetrical problems requires dealing with some quite frustrating differential equations. Great question. My answer would involve so many things in the music space. These are mostly actually impossible right now, but being able to think about things that you want that should be possible and seem completely out of reach is a great technique for finding motivation. The world is much, much, much bigger and more full than our little brains with our little problems can usually understand. So, some of mine: 1. Live, remote, group music performance. Yeah, latency, blah blah blah. This problem can be solved if you have... 2. Music generation as the standard way music is distributed. Some people love hearing the exact same recorded song played the exact same way over and over again. It bores me. And "live" performances are often musically and sonically inferior. Every musician I know is creative, every note uniquely created and delivered. We should convey that uniqueness in the distribution of the music. Yeah, there are tons of "music that sounds like" projects. Have it be part of the musician's workflow. Get them to trust it. 3. Live noise cancellation for spaces. People love noise cancellation for in ear or over ear headphones, where you have tight control over the sound path and the acoustics. Child's play. Solve it for spaces where you may have undesirable sounds coming from any direction, with unusual acoustics and strange surfaces. A $1000 device that does this makes you $B, because it increases the property value of many spaces by 2 orders of magnitude. 4. Accurate performance reproduction. JFC the state of musical reproduction is so completely absolutely shitty. The chills one experiences being proximate to a virtuoso instrumentalist...even the best tube amp speaker set up does not fool a close listener for more than a few seconds. Is there a Turing-test like name for this? There should be. 5. Instrument/track extraction. We're getting better at this, it's the impossible problem we're closest to solving. But there's still a lot to do. 6. Instrument acquisition. Everyone should be able to acquire skills in an instrument. With all respect to the 10,000 hours theory, we are in the dark ages when it comes to acquisition and skill accumulation. Teaching is terrible and pretty much everybody practices terribly. People learn from watching and doing and participating- we are visual copy paste monkeys. With some combination of robots and visual production and feedback systems and nutrition it should be possible to develop benchmarks for dramatically improved acquisition. Cheers, great question. Those are superb points! You might want to know that (4) has been IMO solved (long time audio enthusiast here, spent more than 15 years on solving that exact problem, developed my own startup that was focused on making DACs with the best possible analog sections - I gave up to a better idea done by someone else, which I'm going to mention next. Rob Watts, a DAC designer from the UK has solved the problem with his FPGA based DACs and pulse array analog sections. Instruments' transients and how they impact a listener (human hearing is far, far more complicated than vision) seems to be the key for the brain to mark a given sound a "natural one" and properly place it into 3D space. Watts works exclusively for Chord Electronics nowadays, but he'd started his own kind of DACs in the 80s. The guy is a genius, I have nothing more to say. My long time quest for properly sounding audio source has been finished. Currently he's working on his Davina project which is going to bring the same technology for audio reproduction (and rebuilding) to studios, leveling up his game even more. Stay tuned, because it's going to take some time. The only other company that does something similar is DCS, but they're extremely expensive and they're rather on the fun side of listening than accuracy. Myself, I use Chord Electronics Hugo 2 DAC (that's the cheapest one having all important Watts' technology, ca. $2500; there's also Mojo - very cheap, but it's mostly for the on-the-go listening) paired with either Audioquest NightOwl headphones or my valve custom made stereo. It's a bliss. PS: Regarding (3) there's no way noise cancellation technologies can produce superb audio. Not possible (unfortunately) at this point of technological advancement of how speakers are being build (and close to nothing has changed during the last 60 years in that area), so they should rather be saved for dealing with unpleasantness of very loud environments, not necessarily suited for music listening. Thank you very much, am very interested to learn more. >Live noise cancellation for spaces. I was thinking about if this being weaponized to suppress speech. If an array could be put on top of a protest to silence it. It's sort of the equivalent of shouting over someone. Would the supreme court find the government pumping out anti-noise a violation of rights? A bot that uses ML to trawl through all your social media accounts, get a good sense of what you would post, and then starts auto-generating posts for each platform, for you to review. You look at them once a day, select the ones you want to publish or just click select all, and it publishes them, appropriately spaced out over a 24 hour period. You also need to setup the other side and train another network to like and RT tweets. And use the RTs you get and use them as feedback. A web map that is updated in near-real time. I'm working on a launching a startup (notasatellite.com) that uses a network of cameras on 80,000+ commercial flights to build the most accurate map. Had the idea flying home one day and realized the image quality at 30,000ft is just about as good as a satellite at 300 miles. There are thousands of planes flying around interesting places every day, so 100x the revisit rate w/o needing to spend millions building, launching and maintaining satellites. Original use case was detecting port container volume over 5 minute periods on approach to PDX which proved the feasibility. I've got ~250 customer commitments with letters of intent so far but need help with some of the ingestion pipeline work. If you have GIS experience in any capacity (as an analyst, developer, enthusiast, etc.) I'd love to chat. Are you comfortable talking about the existing pipeline on HN? Do the cameras geo-tag the images into a GEOTIFF or similar at the edge? What sort of spatial resolution do you get on the resulting maps? Any chance of multi-spectral imagery in the future? would be amazing for doing remote-sensing projects Sounds like a awesome project. Thanks! I can definitely elaborate on some of the strategy here. The cameras geo-tag the images at the edge and verify against the flight path using time stamps for accuracy. Resolution is a function of altitude (among other things), but tests have consistently demonstrated <1m resolution at common cruise (35,000ft) and <10cm below 4,000 AGL. Multi-spectral is planned, and the final the prototype is equipped with an IR dome among the other sensors in the array. The entire device is mounted internally on the window which sidesteps the need for FAA clearance or tunnel testing. Awesome stuff, This sort of data is going to be key for future planning/remote sensing work so great to see an alternative to satellite imagery popping up. What sort of issues are you having with the ingestion pipeline? I have some amount of GIS/mapping experience and happy to weigh in if I can. Great search engine for an eBook collection. Gives higher ranking to sections i annotated. Must be a desktop app. Have you tried Calibre? Probably, it's been a long time since. The best i've found is recoll: https://www.lesbonscomptes.com/recoll/ But search quality isn't good. I think they're looking for something that searches content. Calbre search only covers metadata. Need to clarify re ebook format though - if it's DRMed then any third-party tool is going to be out of luck. Someone was working on a full text search plugin for calibre https://github.com/idealist1508/Recoll-Full-Text-Search-For-... I intended to try it out but never did - might give it a look again soon. How about the YC request for startups list? https://www.ycombinator.com/rfs/ An agent that can read scientific papers and implement code for them. Or a much simpler one - an Agent that can read rules for an arbitrary boardgame (PDF) And generate a playable version for you. A user interface that combines eye tracking, voice, keyboard, and mouse input. I'm not sure what might work, but as something to explore/test: I look at a button and say "choose", and the button gets clicked. Or maybe I blink twice while looking at the button to click it. Or I say 'switch to Chrome' and it does so immediately, not after 10 seconds for a Siri round trip to Apple. Eye gestures. Look at button, look left, look back at button. Radial menus, like Maya. Look at button, menu appears, look in direction to open next menu, new radial menu appears. A P2P file sharing application based on Publius technology. Allows anonymous uploads, splits and encrypts files redundantly across servers. The hosts don't know what they're holding. Privacy enhancing and censorship-resistant. Why do we only serve user interfaces over the web via HTML/JS interpreter inside the browser? If you can somehow provide some isolation (something docker like) and not yet thinking about delta changes, why would we not able to load e.g. a native QT application over the network (i.e. using DNS but instead of serving HTML over the webserver, serve a native application)? You can. Flutter does this by making the app run in a canvas element with WASM. Look cool, but I was envisioning a more general approach I suppose, not limited to a single specific language/tool. Flutter still outputs a x.js which is interpreted by V8 and runs within the Chrome sandbox, what I meant was more along the line of a supercharged browser able to run e.g. the JRE allowing to run any random java program out there, but then accessed over an URL and requiring no manual "installation"/copying/extracting files or interpreter setup. Same for e.g. a native Dart app. So you would be able to run any native application (given it's built for your platform obviously) within this kind of browser, you could run chrome in it and access the web like usual I suppose? So I guess just hassle-free native apps with a focus on client/server architecture like websites, running gimp/paint doesn't seem like a good fit for something like this :') Looks like you're looking for WASM apps then. Native programs including games if they're compiled to WASM can run on the browser. Ok WASM looks pretty epic. Google Earth browser version works pretty well and this QT example feels pretty native qt.io/web-assembly-example-slate Any idea of how popular WASM is? I don't feel like I've ever interacted with any WASM app in the wild which seems pretty sad given the possibilities and how bloated websites in general are. Still in development, basically. It's not used for full GUI apps necessarily, more like to run computations in the background that are then sent back to the HTML/CSS/JS based GUI. I see. But theoretically say I want to make a web-accessible GUI I could just make it using my favorite native GUI library which supports compiling to WASM and it _should_ work from there cross platform + mobile, right? Thanks for the chat :-) Accessible VR. I mean monetary and technically. VR is good but the main problems are the price (should be much much cheaper) and you also need a pretty beefy PC. I know it's a new technology breaking in slowly but I really have high hopes as I look back at smartphones for example (just compare 2010 vs 2020). Yet it's still far away to be in every household. I'll throw in a possible solution. A real time VR engine. The industry has spent decades building a graphics pipelines for full fidelity and variable frame rate through buffering. We could go back to something like early consoles that ran at full framerate by displaying a max amount of sprites per frame. Seems worth experimenting with. Neither my idea nor do I think it is impossible to create, but I would love for this tool to exist and don't have the skills yet to make it myself:
http://worrydream.com/#!/DrawingDynamicVisualizationsTalk You can impliment my catastrophy bonds idea[0] [0]: https://asindu.drileba.capital/2020/02/fighting-of-disease-p... Forecasting game (inspired by the book Superforecasters) Wondering if there'd be enough demand for this.. A reliable fake news detector that my older relatives can use easily, and maybe more importantly, that they feel inclined to use instead of just believing and sharing every piece of information they receive. It wouldn't be perfect but something that followed everything back to the original source by following links in articles and looked for loops might go a long way. It already exists. If it's displayed on a television or the internet you can assume it is untrue. I generally accept only first-hand accounts and original sources for information. Not impossible but hard:
create an application that lets you talk and auto-generate code/pages for an app.
There are lots of companies working on this but you might be able to find a good niche. I can't believe we still don't have a standard tool that lets a user design a CRUD application right from server back-end to front-end. I think they do exist. Ruby on Rails and similar for example. Microsoft Access and similar is a bit older but does that too I guess. Microsoft Access is decades old and no one seems to have improved on that. Microsoft FrontPage was also on same lines. Why we are not abstracting CRUD away like everything else? Is it because engineers are happy creating CRUD apps? It's a mistery to me too. No, because the best way to motivate an engineer is to tell them something is impossible. Making a browser do all the work of an OS or vice versa I think somewhere in between these two options is emacs. A native text editor with VS Code's capabilities. Check out Onivim 2 Fully automated restaurant +1 however making a robot that can chop ingredients and make a simple stir-fry looks quite possible now I want the home game version, Chefbot 9 million. Serious. I don't feel like cooking all the time and end up eating unhealthily. As long as its bumper doesn't get stuck or battery runs out before cleaning up the kitchen and docking at the home charger. There are a lot of complex physical jobs people perform that could be automated with enough enginerding and dedication. A bathroom-cleaning robot/s that can do tile floors, sinks, mirrors, counters, showers, and tubs. Cleaning bathrooms is a really, really sucky chore. Also, yard trimming/cleaning robots/drones more than just lawnmowers. Bonus points for cleaning-up after olive/plum trees and sculpting topiaries. A sink/dishwasher robot combination that can clear plates, separate food waste and optionally compost, load a dishwasher, and "hand"-wash what cannot be dishwashed. Bonus points for clearing the table and putting away clean dishes automatically. A good, automatic gutter cleaner robot and/or an inexpensive, modular, micromesh stainless steel gutter guard that doesn't suck. A quality full-tower/super-tower case (similar to what CaseLabs produced) that actually holds an EATX or Supermicro server board comfortably, can support water cooling with an optional distro plate, can hold 11 3.5" drives, 4 2.5" drives, 2 5.25" bays, 6 200mm fans, 2 double-wide relocated GPUs, and options for RGB, temp monitoring, and fan controls that work under Linux, FreeBSD, hackintosh, and Windows. Options for color and tempered glass windows. Harbor Freight-alike that has better quality, is online-only, no coupons/gimmicks, and ships must faster. Between HF and a McMasters. (HF ecomm, customer service, and shipping are a clusterf.) Smart home integration amongst different brands and complementary products, preferably with open standards or a "Dolby NR"-meets-AWS for smart home APIs. For example, I bought an Awair IAQM only to discover they were dropping all of their integrations, including with their parent company's thermostat, Emerson Sensi. My Filtrete Smart air filters should work with Nest and Emerson. An IAQM should be able to turn on the central air fan if the indoor particle levels are high or if a smart smoke detector detects experiences a false alarm. Also, turnkey smart home integration modules for Arduino, RPi, ESP32, and general electronics. Also, federalizing social networks. I know that Mastodon exists, but the impossible part would be making something like that truly popular. Also, speaking of social networks - giving people the ability to automatically sync their charity contributions to their public social profiles, as in "I'm donating 7% of my income to these causes, and 3% to those" (using some properly designed API from tax authorities) and making it popular enough so that most part of the population is socially nudged to some accepted level of regular charity donations. In my libertarian fantasy, it would gradually replace taxes, but yeah, I'm aware of how idealistic and far-fetched that sounds. Not going to work, unless there is a certain way to have 'one person one account' type of tech, or make account creation very expensive. Today Email is effectively federalized, but in practice, unless you have some major spam technology, your inbox will be filled with spam as soon as your email becomes public. Same thing is going to happen with decentralized social netowrks.