reBoxed: Global People Rankings

I'm experimenting with a new feature for reBoxed, the email prioritization tool I built two weeks ago.

By having you and others vote on people's relative importance, reBoxed re-sorts your inbox according to the importance of senders in your inbox. But wouldn't it be interesting to see the most important contacts based on the votes of you and others? That's why I built a feature called "Global Ranking", which gives you the global importance of people in your inbox based on everyone's votes.



I'm enforcing two rules to protect everyone's privacy:

1. You can only see the global importance of people you have communicated with from the Gmail account you're using.
2. You can only see the global importance of people whom you've voted on yourself (either for or against). If you don't see enough people in your global ranks, you should re-play the reBoxed voting game a couple of times.

Additionally, Global Rankings will not show rankings for addresses you've marked as "not a person" - bulk senders, spammers, and the like - that adds a little more sanity to the page.

You can access the feature through your reBoxed Inbox, by clicking on "Global Rankings".

Give it a spin and let me know what you think! As always, you should report bugs on the reBoxed GetSatisfaction page.

48 Hours of reBoxed

Another quick update on reBoxed. I launched it a little more than 48 hours ago. Here's the visitor curve so far from Google Analytics - looks a lot like a hockeystick:



And this is what has happened since my last update:

• reBoxed was covered in ReadWriteWeb and Lifehacker.
• reBoxed now has 414 users - compare that to 128 users at this hour yesterday.

I expect that growth will fall off over the weekend, as fewer people use the Internet on weekends. I've decided that I'll take a few more days trying to incorporate user feedback and make the site more useful (rather than working on reMail's core product), but I'm taking part of the weekend off for my own sanity. Hopefully, you'll see new features on reBoxed sometime early next week.

reBoxed on ReadWriteWeb

Marshall Kirkpatrick writes on ReadWriteWeb:

Former Gmail engineer Gabor Cselle has been working on improving email for years. This week he built a new system for prioritizing all the emails in your inbox. It's called ReBoxed, and it relies on crowdsourced A/B preference voting on email senders, and Cselle built it in just 3 days.

Marshall points out some flaws in our current ranking algorithm (There any many, considering I wrote it so quickly), but closes with this:

ReBoxed is a small project that presumably could become a part of Cselle's larger email startup company, ReMail. That service is yet to launch but if this is the kind of creativity that will be included there, we're excited to see it.

I'll be working on reBoxed all day today and will keep you updated. Meanwhile, check out reboxed.remail.com.

24 Hours of reBoxed

I launched reBoxed a little more than 24 hours ago, at 5 p.m. yesterday.


Since yesterday:

• 128 people have signed up for reBoxed
• 746 visitors have checked out the site. See the Google Analytics pic above.
• I've made 34 checkins to our source code with bug fixes and new features

We've had some pretty good feedback so far, and users have also tweeted me with suggestions and bug reports - I've set up a GetSatisfaction site for the latter.

reBoxed also now has a "Delete all my data from reBoxed" feature. You can use that to delete all of reBoxed's knowledge of your contacts and email. You can find it at the bottom of your reBoxed Inbox:


Next, I'll be working on the reBoxed homepage. I want to make it more attractive, and I'll break it up into "Home" and "Learn More". I should have that done by tomorrow.

Should reBoxed have a "Skip" button?

reBoxed is growing - people are signing up! Pretty exciting to run the script that spits out the number of users. I'll give you an update on user numbers tomorrow (after I've caught up with sleep).

reBoxed lets you vote on your contacts and reorganizes your inbox based on your and everyone else's votes. This is what that looks like:


@andyy writes on Twitter:

@gabor repeatedly wanted a "can't decide/equally important" option - am I too apathetic? :) [...]

I think that reBoxed should not have a "Skip" button. I want you to make a decision. Those two friends that you both value? Your sister vs. your brother? Make a choice. This might sound radical, but one of those people usually does write more important emails.

reBoxed is up!

Try it now at:

reboxed.remail.com

reBoxed sorts unread emails in your Gmail Inbox by the importance of the sender. You start by voting on pairs of contacts - "whose emails are more important?". Your votes are then combined with your friends' votes. reBoxed's algorithm then sorts your inbox based on this input.

reBoxed is your inbox, sorted by your friends.

And we built all of it in just 75 hours. (Well, more like 77 - I needed an extra 2 hours to iron out some kinks at the end).

reBoxed - Almost done

It's 3 pm on Wednesday, which was my personal deadline for reBoxed. It's not quite there yet - still working out some kinks. But it should be up in the next hour or so.

reBoxed - Day 3

I woke up to the sound of the alarm at 9:00 am this morning, and rushed to pick up a friend at SFO. Should've thought about that when I went to bed at 4:30 am ...

I only spent about 5 hours coding on reBoxed today, mostly fixing bugs and making the UI at least bearable. We spent hours discussing derivative ideas, and probably got a little too excited about the eventual potential of reBoxed. Then, we met with Paul Graham at YC office hours - he seemed optimistic about the idea, but pointed out some UI shortcomings that are going to take a lot of effort to improve.

He had one more point: reBoxed will require quite a bit of input from the user before reorganizing their inbox. I hope users will understand that we can't just reorg your inbox out of thin air, and won't get too impatient with the work we ask them to do.

The quality and speed of the UI, however, will continue to haunt me. I feel like I'm pretty good at designing and implementing UIs, but making them good takes takes enormous amounts of time. With my deadline at 6 pm tomorrow, there's just no time get to get the UI up to my personal quality standards.

Plenty of small work items remain, but I think we're on track to push the first version tomorrow Wednesday at 6 p.m. - Exciting!

--

P.S.: If you're wondering why I only coded 5 hours new today - The Winter 2009 YC founders decided to continue the Tuesday dinners even past the official YC dinners. Tonight, we had a little get-together at Cloudkick HQ in San Jose. I just got back from that.

P.P.S.: I just noticed my original post said I'd have V1 up by Wednesday 3 pm - So that's what I will aim for (instead of 6 pm). I'd say 6 pm is more realistic though.

reBoxed - Day 2

Wow, it's 4:10 am - I started working on reBoxed at 10:30 am today. Almost 18 hours and 1045 lines of code later, I have a crude prototype working, in line with my plan from yesterday.

reBoxed still seems like an excellent idea, but I've found that it's crucial to keep expectations down - since the reBoxed idea is so new, I still find myself being wildly enthusiastic about it. I typically get much calmer about a new idea after about a week or so.

The first version is going to suck. Especially the UI. I'm going to spend some time sprucing up the user interface on Tuesday, but I doubt it's going to look really good in the next 36 hours.

I wasted some time today by deciding to look into fancy machine learning algorithms for reBoxed. I dusted off my copy of Data Mining (the only ML book I own that isn't with my parents in Switzerland). But after 45 minutes or so, I abandoned all of that and went with something really simple.

People have emailed and asked reBoxed - it's a combination of a number of ideas, some of which I've written about in A Model of your Inbox and How Researchers are Reinventing your Email Client. You'll find out soon enough.

Ok, off to sleep now.

reBoxed - Day 1

I'm working on an idea called reBoxed and want to launch the product by Wednesday night.

Day 1 went pretty well. I spent an hour setting up a test server, and then most of my time was spent learning Oauth, which I'm going to need to use for getting access to some data I need. I decided to use Leah Culver's Python Oauth code and adapt it for my purposes. I was going to write a Design Doc but decided to just scribble something down for lack of time. This will likely come back and bite me. I was up until 3 am last night and started working at 10:30 am this morning.

So far, on target for Wednesday.

reBoxed - An Experiment in Rapid Development

I have a little bit of free time on reMail's main product right now and I've decided to implement a possibly great idea I've been thinking about over the last couple of weeks. reBoxed is a new take on inbox prioritization - and I'll build and launch it in the next 3 days - 75 hours. My deadline is to have it done by Wednesday 3:00 pm Pacific time.

My timeline is something like:

1. Today Sunday: Learn a technology I'll need to understand for reBoxed to work. Write a Design Doc.
2. Monday: Implement all functionality but keep the UI super crude.
3. Tuesday: Brush up the UI. Meet with PG at office hours in the afternoon to get his feedback.
4. Wednesday: Write deployment tools and deploy to a server. Launch.

Don't set your expectations too high: The first version of reBoxed will likely suck. reBoxed is an experiment in how quickly I can develop and launch a mail-related product. And a way to prove out a possibly great idea.

I'll post a daily update on this blog.