Projects that need to be built.
Problem: Finding something good to eat.
Yelp’s quick-tips + foodspotting (more focused version of Oink)
Stopped working on this idea when I heard of Oink, figured Kevin Rose can do better.
Oink didn’t seem to go anywhere and Yelp isn’t innovating much, so this is still worth building.
Problem: Expense of setting up error-handling and updating framework for games/apps.
EC2 + S3 based Steam-clone oriented towards developers. Ideal for testing new applications and games. On crash collect minidump and return an analysed stacktrace/thread-deadlock-analysis/etc to the developer.
Developer can save several months of effort by not implementing: auto-uploading crash-reporter, process-dump-analyzer, auto-self-updater, online file host. Can also provide basic analytics, and a way to notify users of updates and gather feedback.
A nice future feature would be an in-game bug-reporting overlay with screenshot functionality, for those pesky bugs that need to be seen to be understood.
Prototyped it with an rsync algorithm on EC2 (without S3/CDN), and WCF/WPF based server/clients. Stopped cause didn’t have branding/design.
Problem: Difficulty of choosing a good movie to watch tonight.
Netflix, Freebase, and RottenTomatoes. Unforunately all three have different web-service APIs, making it a pain in the ass to integrate. Freebase has IMDB->Netflix MovieId-reference entries, and RottenTomatoes comes with IMDB MovieIDs. So I can go RottenTomatoes<->Netflix with the help of Freebase. Make a nice XBMC-like UI and put it on a tablet.
Incremental video codec
Problem: Making small adjustments to videos through editing/compositing is extremely painful. There is no good online video composting collaboration supporting technology. Sending losslessly compressed video over the web is too slow.
Deterministic compression which can be broken down into blocks which can be synchronized over a slow-link via some kinda delta-compression might help.
Use case: a mistake on your title screen require only re-rendering of those first 3 seconds of your music video. and then fast sync against the web-hosted video over a typically slow american internet connection (1mbit up)
Compressed HD video with Alpha codec
Problem: There are no accessible compressed/efficient video codecs with good alpha support.
At least as compressed as HUFFYUV. More compatible than Quicktime Animations. Work with WMV9 surfaces, in DirectShow, and on Linux (V4L?). There is a serious lack of immediately-playable video codecs, and an even bigger lack of compressed codecs with alpha.
Realtime HD video playout
Problem: Lack of low-cost, stable, efficient HD video play-out systems in TV production.
From what I’ve seen, it’s very painful to playback a video clip, particularly an HD clip. Seeking a file from disk, pulling over network, buffering a couple of frames, decoded into memory, uploading decoded frame to gpu, copying over hd-sdi, compositing into a scene; all adds up to delays that require a lot of logic to work around in common production broadcast video systems. And yet its completely doable. Just requires lots of painfully written code. Lots.
Problem: Incomplete downloads from unreliable multi-part hosts.
Determine which articles must be redownloaded to complete a file from a newsgroup, and look for those articles on alternative servers. Could break this out into a separate web-service which scans multiple servers for missing articles to determine the best newshost. A lot of effort for little reward, much easier to just redownload something using a client that supports multiple servers.
Single stream chat multiplexer
Problem: Duplicate msgs on your phone, desktop, iPad when someone msgs you over FB/AIM/etc. There is no way to dismiss the msg from other devices when it is handled on just one. Have to unlock your phone and dismiss it there, close it on your iPad, etc.
Use case: A message hits my phone, then sits there for 5 minutes without being spotted. It forwards itself to my work desktop or my tv, where I acknowledge it. Message may remove it’s notification from all other systems, so that I don’t have to go to every device and discard it manually.
Pie based menus
Problem: No one is doing pie-based menus. Seriously. This is the best thing ever for touch-screens.
Build some simple apps with extremely clean UI that use pie-menus effectively.
Research: Monte Carlo based video codec
Noise looks better than predictable motion compression artifacts. See if there is compression that may hide its artifacts in grain rather than block motion vectors. Plus “grainy” seems to be in style these days.
Compress a geographic map to only show points of interest with relative not-to-scale distances between items.
Use case: shrink a giant map of Manhattan to display the 10 restaurants you may want to visit this Friday night. Isometric pixel-art style would be perfect. If only I had a good pixel artist to work with.
Problem: We really have no idea what is going on in our own city outside of the spotty coverage you get on TV.
Photos from around the area. What happened in Lower East Side last weekend? What did Park Slope look like during Snowmageddon? What is happening in that park with the cops pepper-spraying protestors right now?
Prototyped the system with free data from zillow, easy point-in-polygon testing of neighborhood shapes sql-based GIS, upload endpoints that mobile phones can hit, GPS support. Need an artist to design a compelling UX.
Anonymous Reputation System
Privacywall that vouches for a user being an individual and not a troll or bot. The user may only have one account — they will want to hang on to it. You plug real information into one side of the wall, and on the opposite side of the wall you get a unique one-way hash that represents you in an anonymous way. This is a web-service others can use in place of oauth/facebook-connect/twitter-connect.
Developers don’t have to implement complicated security and user-accounts. They can allow anonymity without spam. If a user misbehaves, she only has one account, and that account may be banned.
You can increase your “account validity” by connecting facebook/twitter to your account, passing captcha tests, receiving a post-card at your home address, even using your credit-card. Since its really hard to commit fraud across all these systems at once, this lets us validate accounts really well. Saving developers months of work. At the same time, the Privacywall lets people speak their minds without risking their careers/social-lives by never releasing any of this private information. So you appear anonymously on any website, but have a reputation to maintain anyway.
Optional feature: UserIds are not traceable and are unique per domain. So multiple websites can’t tell if they have the same person on it. Or website A can’t tell website B to ban a user. However they can hurt the user’s reputation. All without sharing any private information.
An economy could be used to throttle any website that is harsh on peoples reputations. So the effects of any domain on your rep could be normalized using information like “how often this domain submits negative rep reports compared to other domains”
I don’t see a way to pay rent with this, so skipping it for now. But I really want this to exist and to be operated by someone trustworthy with a lot of transparency. Ideally such a system would keep your information safe even in a complete leak of the entire database: any private info need only be calculated once then discarded, and references inside the database could be one-way hashed with some serious salting. (CryptoDB?)
Public Bug Tracking [WIP]
Problem: App Marketplaces are mixing reviews with bugs in an ugly way. A perfectly good app can get bad ratings because of a few bugs, which will bury the app and keep people from discovering it. If an app crashes on startup, it will receive a streak of 1-ratings that will bury it on launch, from which you won’t recover. Meanwhile the app may be excellent, but have a bug that causes crashes on a particular phone. Considering that most developers can’t make money from phone-apps, this is a big deal.
We need to separate bugs from an app’s overall rating.
Id like a website that lets you submit bugs to apps, games, websites etc. This is a more main-stream version of the responsible-security-vulnerability-submission idea. It should also help show the people who build marketplaces that a beta-market and bug-support must be first-class citizens. Looking at you, every app marketplace. (Nokia has a beta market, but its obscure.. so no one knows about it.)
This could also provide the go-to source for requesting feature improvements, voting on features, voting on getting the most important bugs fixed. Developers would love to know where to invest their time. And users would love to get their moneys worth by actually getting fixes and updates that THEY want. Win Win!