Why I think Howstr will change the world

Howstr is a system that makes it really easy to find out how to do things. In simple terms, Howstr considers writing and reading overrated. They’re great for sharing stories, but not great for getting work done.

what are books good for? Reading: yes. Flying a plane: no.

Howstr replaces a simple story about a project with a complex model of what an expert knows about the project. Knowledge, not words. (okay, it uses words, but they’re used better)

The expert explains everything to Howstr. Howstr explains it to everyone else.

Continue reading Why I think Howstr will change the world

The Forest And The Trees And The Leaves And So On

We think in networks but we have to communicate in sentences. Sometimes we can sketch out a diagram or find a stock photo, which saves a bunch of words.

But even using multimedia doesn’t seem to save us time in the long run. Whatever we produce still only works for a certain audience and it still falls out of date. We keep pushing the rock up the hill just to have it roll back down. As if that wasn’t enough, there’s always the lingering feeling that someone else has probably already explained this if only we could find it.

What we’re dealing with is the cost of getting thoughts in and out of our heads. The conversion doesn’t come cheap. It’s like trying to live-tweet through Burning Man using Morse Code.

I think we can solve most of this problem if we can keep information related to itself while it’s outside of our head.

Continue reading The Forest And The Trees And The Leaves And So On

In Which I Get Excited, And Then Disappointed, By Sam Altman’s Latest Post in The Macro

I say that the stuff Y Combinator (and the people associated with it) has put out over the years is the core of my startup education. There’s always something valuable in what they share.

However, this one has me a bit confused. Sam Altman (Y Combinator) asked Michael Moritz (Sequoia Capital) some vague question about why Sequoia is so dope. Then Michael answers. Then Sam says “that was an incredible answer.” What happened in between was apparently so great it was pulled out of the normally private YC dinners and fixed in text for all time.

It stands to reason that both of these people know what they’re talking about, and that this excerpt must be the best bit of them talking about what they know. So, logically, diving into this transcript should be more valuable than most other things I could spend my time on.

Continue reading In Which I Get Excited, And Then Disappointed, By Sam Altman’s Latest Post in The Macro

The Dark Side of Frictionless Digital Sharing

The general principle behind why services like Uber are so successful is that they commoditize things. Getting someone to drive you around used to be hard because only a few parties could do it. Uber allows pretty much anyone to do it; Uber drivers are interchangeable. The system just has a simple formula:

X passengers + Y start + Z end = $

The system doesn’t care how that equation gets balanced. Right now Uber is using independent drivers. Soon they’ll use driverless vehicles. Someday they might use toll zip-lines. Whatever. The success comes from the scale of allowing every possible solution to connect to every possible customer.

So, that’s cool, but the tool itself is amoral. The uses that people put it to can just as easily be wrong, or unethical, or at least off-putting as they can be heroic or pedestrian (so to speak).

And example of the more questionable activity enabled by platforms like this is that sites devoted to “tutoring” make it easy to connect people who want to cheat with people who will do the cheating. By commoditizing the purchase of temporary expert knowledge the system can’t help but attract cheaters. The interchangeability is enabled by the testing platforms themselves which don’t care how the problems get answered, only that somebody answers them. So students can just hand over their login/password and have a hired gun do the work for them.

These platforms can produce a lot of value because they eliminate uncertainty and friction, but in so doing they also have to give up security. Verifying that rules and standards are being adhered to is inherently intrusive and slows things down. For example, when the Uber-for-packages services get established there’s no way they’re going to be able to care what’s actually inside the packages.

Automatic Assembly/Dis-assembly Precedence

When you get done designing a beautifully complex system you have that moment when all of a sudden somebody has to actually build it. I don’t have any data, but I’m sure people have died over manufacturability.

This automated assembly and/or disassembly stuff takes the designed system and tries to let the computer figure out what order to build it in. For example, an algorithm will basically “wiggle” all the pieces in the 3D model and see which of them come off without being blocked by another piece. Those go into the “first” level. Then wiggle all the remaining pieces and see which of them come loose; put those in the “second” level. Repeat.

That process gives you relations like (Part A [came off of] Part B). Doing it automatically requires a really good 3D model and doing it fast requires a really good computer to manipulate the 3D model. Even then, it’s common to make unrealistic simplifying assumptions like pretending every part is rigid. But it works for most things.

Improving Assembly Precedence Constraint Generation by Utilizing Motion Planning and Part Interaction Clusters

Real-time, Collaborative, Heterogeneous CAD modeling

There are a lot of different CAD packages which don’t normally work together. However, the actual modeling is basically the same. Putting a hole in something is putting a hole in something no matter what the database looks like.

So apparently it’s possible to describe (nearly) all of the operations that CAD modeling programs use to do the modeling in a neutral language. Stick an appropriate plugin onto your favorite CAD modeling program and it will translate your actions into the neutral language, send a message to everyone else, and a plugin on their favorite CAD modeling program(s) will execute the same operation in real-time.

As long as you’re okay taking turns you can all work on the same model, from different places, using different programs.



Real-Time Collaborative Design With Heterogeneous CAD Systems Based on Neutral Modeling Commands