August 23, 2013

4,000 Mile Brainstorm

My wife’s new job at CCP in Atlanta came with a relocation package for all of our stuff, but not our cars. I drove my car out in July (via Tombstone, White Sands, and Carlsbad Caverns) to scout for housing and daycare, and we decided to turn her car into a family road trip. To spare the baby long-haul hours on the road, I drove her car from Redwood City to Kalispell, MT. After picking the family up at the airport, we started a slow crawl across Montana, Wyoming, and South Dakota–Glacier National Park for three days, the Rimrocks, Bighorn Canyon, Devil’s Tower, Mt. Rushmore, and the Badlands. At that point I dropped them off at an airport and long-hauled it down to Atlanta.

I love long road trips because your mind can really wander, and on this trip I decided to keep a road journal of sorts–all the wacky business/digital/game things that I came up with.

(1) AirBnB for Tour Guides

This one was actually a little older (thought of it on road trip #1). My friend and co-pilot on the first road trip is under-employed in L.A. (aka making a go of acting), and we were trying to think up things that would make him money. He often takes friends around L.A. and regales them with his encyclopedic movie knowledge, and it occurred to me that someone could make a marketplace for such tours. Allow anyone to sign up for “J.R.’s tour of Hollywood”, take 10%, and call it a business.

Like AirBnB (at least when they started, not sure if they still do), send a professional photographer on every “trip” and vet them for quality along with adding some slick photos. The biggest cost would be finding and creating the initial offering catalogue, so it might make sense to stick to a few destination cities before opening it up to anywhere (why not a tour through the middle of nowhere in Montana?).

Someone later pointed me to Vayable, which is pretty much the same thing but limited to only a few cities… and seemingly focused on exotic destinations for now.

(2) Geared, a surrealish iPhone game

The one liner in my notes: “You are a Blade Runner style forensic robot detective who puts dismembered robots back together enough to ask what ‘killed’ them.”

I sketched out a really rough mechanics diagram for robot repair.

The basic idea is that you’d be looking at something like a clock face, with 2-4 different arms on it. There would be a different gearing ratio between each of the arms, such that when you grab and spin one of them all the others would also rotate (at different rates).

After sketching it out, though, it occurred to me that if the gearing ratios are fixed you’d basically be able to just grab one arm and spin it until the puzzle was solved. It would take some prototyping and testing to see if this is even something that would work as a puzzle mechanic.

If not, another puzzle mechanic could work–the game is really about content more than it’s about a specific puzzle mechanic. I’d probably have to create 10-20 different models of robots in various states of disrepair. Each “case” would basically have a picture of a dismembered robot and 1-3 “clues”, where each clue is one of the mechanical puzzles above. Solving all of the clues would reveal what finally happened to the robot, in a sort of dialogue-story mode.

Would there be a story arc? Not sure. Most likely the “cases” would be seeded with a ton of content and then procedurally assembled so that you get a sort of existential game where you do your job and that’s it. There are always more robot crimes to solve!

I like the “world” of this idea a lot, but ultimately it’s probably a little too content-intensive for me to take on right now. It’s mostly a writing exercise, along with a crapton of robot sketches and UI work. I may take a crack at prototyping some little clockwork puzzles at some point, but I can’t see it making back the $5-$10k or so worth of art that the game would require.

(3) One Rec – One Recommendation at a Time

This is kind of a re-visit of an idea I had a few years back, another way for my actor friend to make some cash. The recommendation engine for things like Amazon and Netflix are pretty good, but never as good as actual human recommendations from your friends. “Oh, you liked XYZ movie? The actor who played this minor bit character in it was in this other movie that’s awesome…” — and sometimes not even a friend. Think about when you used to actually have to go to a store to rent movies. The guys and gals who worked behind the counters at places like Blockbuster were treasure troves of movie recommendations. Now that digital is taking over, we’ve lost those personal recommendations.

The idea came back to me because I’m seeing a lot of those same problems with app discovery and other content verticals. What do I watch next? What do I play next? What do I read next?

I think it would be cool to build a recommendation service that offers one high-quality (i.e. human-generated) recommendation at a time, with the assumption that the user is going to trust the recommendation and go give it a shot. Instead of listing ten or twenty shotgun-style recommendations based on digital filters and hoping the cover art grabs someone’s attention on one of them, speak to the customer like a human being. “You like a lot of James Cameron movies, but you haven’t seen the Abyss? Go watch that. Right now. Report back when you do.”

Customers would report back with four options:

  • (+3) GREAT recommendation (I enjoyed it greatly)
  • (+2) Good Recommendation (I can see why you recommended it and it was pretty good, but I didn’t LOVE it)
  • (+1) Bad Recommendation (I didn’t enjoy it, but I can see why you recommended it)
  • (-3) WTF (Why the hell did you recommend Terminator 2 after I said I liked Sideways?)
  • Skip / Already Watched (these would have to be rate limited to not take up too much recommender time)

Those numbers on the side would be a rating scale, allowing the recommender to earn points in that content vertical (movies, apps, books). At the end of each pay cycle a certain chunk of the revenue would be split up based on how many points each of your reviewers racked up. Great reviews (+3) would create affinities between customers and reviewers so that reviewers who’ve given great recommendations in the past are more likely to be paired up with the same customer. Different verticals would pay differently (people consume movies faster than books, I assume).

This would have to be a subscription business model — probably something on the order of $2/mo. per category or $5-$10/mo. for all categories. The platform (thinking Apple here) would take 30%, leaving 35% to split between operations/customer-acquisition and 35% to pay out to your “Taste Agents.”

The technical part of this site wouldn’t be too difficult (making it subscription will limit the number of customers). The hard part would be customer acquisition and making the economics work. Assuming a very low goal of paying out a “full time” Taste Agent $30k (a low, but not awful, salary for recommending movies), you need to be able to pay them $2500/month. If the “agent pool” is only 35% of revenue, that means you need about $7200 in revenue per full time agent… 3,600 subscribers. Can one person handle the requests of 3,600 individuals? Assuming they don’t all get a recommendation every day (maybe one a week on average), that’s around 500 recommendations a day, or roughly one per minute for eight hours straight. If we hike the price up to $10 a month (too high in my mind), the numbers get a little better: 720 customers per agent, roughly 100/day, roughly one every five minutes. Five minutes is certainly enough time (maybe too much) to browse someone’s history, read their response to the last recommendation, and formulate a new one. I’m just not sure you could convince enough people to pay $10/mo. for such a service.

(4) Travel Beacon

I actually like this idea a lot. Organizing trips is a giant pain in the ass. My wife and I are typically the organizers for our friends, and we try to do a multi-day national park trip once a year. Even with just the two of us, though, the amount of details involved can get pretty crazy — planned driving routes, where we’re sleeping, which hikes we’re doing.

I would love it if there was a tool for organizing this type of information all in one place instead of scattering it across multiple email threads, chat logs, verification emails, google docs, and text files.

The general idea is that you’d set up a Beacon for a specific trip, which would have a private URL. Instead of keeping a marathon email chain going, you’d just send everyone the URL. When they visit, there would be a self-contained discussion thread and a calendar/wiki for entering in trip details. A calendar (in the sense of Google Calendar) is too limited–you only get a tiny box. A full on wiki is a nightmare organizationally… instead picture index cards (trip details) which could be arranged vertically as an itinerary. Each trip detail could have its own discussion section, metadata such as how much that activity costs or how long you think it would take or even things like addresses. A map widget could pull out all the addresses and show them on a map, filling in driving times between activities…. Tools like Trello could do a lot of this stuff, but it would kind of feel like ramming a square peg into a round hole.

It was at this point that I realized I was basically describing Google Wave. I still think it could be awesome, but I haven’t really thought about how you’d turn it into a business. People would probably only use it once or twice a year, so making it subscription based probably wouldn’t work. I like the model of useful desktop software like TextMate or The Hit List where it’s a one-time $50 fee, but the app really needs to live in the cloud (no one will collaborate if they have to install something).

Planning a vacation is basically taking a giant possibility space and whittling it down into a specific itinerary. There might be other sectors that would find that useful (syllabus planning? conference itineraries?), and maybe it would make sense to white-label it for other sites… but it’s a pretty far reach from planning trips.

(5) PartyBeacon / SocialBeacon

This is just a riff on the idea of TravelBeacon. I built an app in my grad school PHP/mySQL class that was basically google+. You added friends to it, and arranged your friends into subgroups. It was all built around the idea of going to pub trivia–I couldn’t very well invite my friends in Athens to a night out in Atlanta, so I wanted a way to just put all my usual drinking buddies into one group. I wanted to be able to send them all a message at once with a time and a place and get a head count for who was down for it. It was ugly as sin and I never actually tried to use it for evening planning, but thinking through TravelBeacon reminded me of it and I wonder if a lot of the same metaphors could be extended to more low key gatherings.

Facebook Events could fill that need, but I have a lot of friends who don’t use Facebook, who rarely get on a chat client, and who only check email at work. The ideal night-out planning tool would have per-user settings for how to contact people (email/fb/chat/sms) and allow people to respond via that channel with a yes/no and maybe a short message. The value here wouldn’t be in the app itself (I imagine lots have tried this idea), but in building the plumbing to get the word out via whatever channel users prefer.

There’s no obvious business model for this idea either, but the chat/sms apps on mobile are growing like crazy, so probably just study how the Asian ones are making money and duplicate that.

(6) The Arcade Girl

This isn’t an idea for a product so much as an idea for a script or a story. It would actually fit quite well with the world of Geared, so maybe I should just combine the two into one story. Words are a lot cheaper in that I don’t have to pay someone for art, but also a lot harder in that I have to sit down and write a bunch.

The crux of the idea is an extension of current skeezy ad models. If you take the phishing schemes and viagra ads of today and extend them into the future, what might they look like? I thought of a guy meeting the perfect girl (manic pixie dream girl) and going on a first date. The date goes very well, with an emotional connection forming. At some point, a waiter pulls him aside. The girl is not real. If he wants the date to go any further, he’ll have to pay. Packages start at one night stand and extend through a one year relationship (best deal!). Also, if he’d prefer a different girl they have a catalogue and can switch out her looks (but keep the same personality) for an extra fee.

There’s a lot of ways you could go with the girl as well. Is she real? Just a con artist? A robot? An AI? What if the body is real but she’s being piloted by someone else? Maybe in the future we can remote-pilot other bodies, so maybe good looking people hard-up for cash rent out their bodies for a few hours each night. Maybe he runs into his girl later, and she’s an embarrassed waitress with no memory of him. She was only being piloted by the equivalent of a futuristic phone sex operator.

I’ve mostly got my wife on board with working part time while I tinker with some video games, but she gave me the “crazy person” look when I pitched her this idea. I don’t see myself working on any sort of straight-fiction any time soon, but it’s fun to think about this kind of stuff on long drives!

(7) Safespeed / Smartspeed

I may have gotten a ticket for going 86 in a 75 somewhere outside of Sheridan, Wyoming.

I remember in college driving up to New Jersey (the good part) to visit a friend. He warned me, something to the effect of: “Once you get to New Jersey, stop speeding. Don’t go ten over like everyone does at a minimum in Georgia.” It would be great if you had a friend like that for every state.

It would be trivial to build a mobile client and backend for people to report the city, speed limit, and speed whenever they got a speeding ticket (getting them to do it would be the hard part). If there were a way to scrape this data from court filings it would be even better.

However you get the data, you could then put that data back on the map (filtering for maybe six months or maybe just the last 20-30 data points per region). Take the lowest ticketed speed, subtract one or two, and BAM! That’s your “safe speed” or “smart speed” that you can reasonably travel without getting a speeding ticket.

It obviously wouldn’t be foolproof and would have to come with disclaimers a mile long, but this is an example of an app that would be pretty easy to build that I would use all the time.

(8) Programming for Housewives (and Plumbers. and Accountants. and Teachers)

One thing I think a lot about while driving through middle-of-nowhere America is how useless computers and technology are for the vast majority of people who don’t live near major population hubs. Somehow that train of thought got me thinking about how cool it would be if programming was an on-call job like plumbing. Like, each town has a few programmers that get routinely called out to solve programming problems. That got me thinking about what that on-call programmer might do on a single call. I couldn’t think of any. As much as it seems like it sometimes in Silicon Valley, most of the U.S. is not very digital.

I turned the idea on it’s head and started thinking about how I could bring about this future programmer-as-plumber scenario. How could a little bit of scripting make the life of a housewife better? A plumber? An accountant? A teacher? When my wife and I were shopping around for an investment property, I used a heavy dose of scripting to automate data collection. Our agent gave us a massive listing of properties–I like to visualize them in a spreadsheet, so I wrote a script that uses the MLS numbers to pull the address, number of bedrooms and bathrooms, rent zestimate, and a few other properties… all into a format I could easily toss onto a Google doc. This saved me hours of data entry, and had my agent jokingly begging me to teach her how to program. (She shouldn’t joke! I’m sure a little ruby would make her job way easier!)

I would bet that there are tons of other professions that a little bit of scripting could help tremendously. I’d love to spend some time shadowing people in random professions and figuring out how scripting could help them automate some of the drudgier tasks. Throw in a couple of basic intro chapters suitable for anyone, then have a different chapter on useful scripts for each profession and call it a book!

(9) Filthbot / Inappropriate Simile Bot

This one is pretty random. I was thinking about how things like Louis C.K.’s standup and Cards Against Humanity have basically made it more acceptable to be vile. I think you could adopt that same mentality and build a really inappropriate Twitter bot. I would just make a giant spreadsheet of horrible similes (starting with Like), such as: “like a rape victim crying in the shower on a Lifetime original movie.” (making fun of actors, not actual rape victims! but still horrible) Then make a simple node or sinatra server to spew these out once or twice a day.

Because twitter displays timelines sequentially, these similes would be randomly jammed up against other peoples’ real and presumably-not-horrible tweets, creating humorous mish-mashes. People could take screenshots of the compositions and respond to the original simile with the two-tweet combination. This would allow people to click on the original tweet and view all the responses (useful, thanks Twitter!). Either me or a friend could browse through the responses and pick favorites to retweet–or automate it. Make it so any response to one of the Filthbot’s tweets that gets above N tweets automatically gets retweeted. Robot humor!

(10) Film Sequence

After pretty much deciding that the economics of OneRec were too rough, I was thinking about other ways to solve content discovery in a streaming world. Another issue that I think about with Netflix is the loss of synchronous viewing. People don’t watch the same things at the same time any more. There’s no concept of channel browsing, and that means every time you turn it on you have to make a decision about what to watch. Sometimes this is hard. Sometimes I spend fifteen minutes trying to figure out what to watch, get frustrated, and then don’t watch anything at all.

What if there was a daily recommendation for what to watch? Programming of a different sort! What if it had some commentary on why that specific movie was chosen? The Discovery Channel does this marvelously with Shark Week–everyone talks about sharks during Shark Week! What if you could do that every week? Just pick a theme, pick seven movies, do a little bit of research and writing, then film a little micro-show about that week’s picks, split into each day’s pick.

People wouldn’t have to watch them in order or at all, of course, but knowing that a lot of other people are watching and talking about a movie makes “keeping up” a little more appealing. How awesome would it be if every time you logged into Netflix, there was a little video playing explaining the theme and the pick of the day? This “show” would cost nearly nothing to produce (an ESPN style set, maybe? a couple of hosts? a couple of writers? cameraperson and director for half a day?), but could probably do quite well selling sponsorships (whatever movie is opening that week, maybe?). I think it would be a fun project.

And that’s it!

That was my 4,000 mile brainstorm. I don’t know if any of these have enough legs to make me run out and start working on them (maybe TravelBeacon), but it’s fun to think through some of the logistics of each project. Until some megacorp offers to bankroll me to just make interesting things, I’ll keep working (slightly) and tinkering on my games that might make some money some day!