s12e13: Exaflops Ahead; The Store with Too Much; Deceptive Design
0.0 Context Setting
It’s Wednesday, 1 June, 2022 in Portland, Oregon and it’s going to be another warm today. The current forecast is a high of 25 Celsius / 77 Fahrenheit, and yesterday we probably ended up hitting a couple degrees higher.
A brief interlude
Did you know I do 1:1 consulting sessions on demand? You should book one!
Glenn Poppe was one of my first customers and said this:
My reaction to the term “thought partner” is usually to throw up in my mouth a bit, but @hondanhon is legit amazing at it.
Dan was quick to grok our biz & challenges, clear-eyed about what matters, and yes-anded to a few new product ideas I’m still pumped about. Go sign up!
1.0 Some Things That Caught My Attention
Exaflops Ahead
I’m thinking about weather. And climate.
A 2004 estimate on weather forecasting and modeling from the computation side predicted that we’d need between 10 Exaflops to 1 Zettaflop for a full global climate simulation2.
In the last week, the Frontier system at Oak Ridge National Laboratory was the first computer to break the exaflop barrier1. So we might have the computational capacity for full global simulation within, what, 10-15 years? (The capacity might exist, whether it’s available for climate simulation is a different issue).
The Store with Too Much
The problem with being an Everything Store is that it would have Everything in it. You might at first think that Everything is fantastic, because you really do want to be able to buy Anything, but it strikes me that there are many different axes that Everything could operate on.
There’s your usual Everything, which would be Everything from clothing to fabric to books to DVDs to washing machines to zip ties to brackets to toys to esoteric hard drives to those fluffy duster replacement things and so on. The ultimate department store.
But then there’s the Other Everything, which these days is the practical experience of “six million HDMI cables” and no way to tell the difference between them. Or even if there is a difference between them.
See, that’s the thing about commodities. I’m not going to condescend and do the “the dictionary defines a commodity as”, but now I’m thinking about this: what point of view does an Everything Store take on selling commodities? If I go to a DIY/hardware store, they don’t show every single version of the same kind of screw. I mean, part of the reason is that they’re limited to the constraints of the physical universe. Atoms take up space.
This is a big question, right? If you’re dedicated to being an Everything Store, then one of your axes is Every HDMI Cable Everywhere All At Once, which includes all colour combinations and lengths and so on, but also Every Black HDMI Cable Everywhere All At Once and Every Gold HDMI Ultra Speed Cable Everywhere All AT Once. Sure you can say “search will fix this” which is not an answer to the problem, which is this: you have set an explicit goal to sell every HDMI cable. In Amazon’s case, they’ve decided to act as a storefront to suppliers, and (it feels to me at least) that there’s minimal effort at dealing with the explosion of choice. I can go to a particular HDMI cable from one manufacturer, go to that product page and if I’m lucky, that product page has length and color and other options for me to pick a sub-SKU.
Look, if I search for “hdmi cable” there are over 3,000 results, each of them have pictures of, well, HDMI cables (and it’s not like I can tell the difference between photography of a physical product versus a render. How does Amazon solve this problem? A bunch of filters on the left, many of which aren’t even applicable to “hdmi cable” (I can choose to filter my search by: aux, coaxial, component, composite, display port, DVI, ethernet, fiber optic, HDMI (of all things), lightning, USB, VGA, XLB), by colour and so on, shipping, new/used, in a design that I’m sure has been tested to destruction and yet also hasn’t changed much since, I don’t know, HTML2. (Which doesn’t mean it’s broken!)
Another thing Amazon does is to prioritize their own brand (the top result), then an ad (a sponsored result), then a best seller, and then an editorial Amazon’s Choice, then a carousel of “highly rated” but also sponsored products, before “more results”.
This is… too many? But also maybe most people don’t care, and they just buy the top result. Or, you know, the product is complicated and you need to get just the right version, like when I accidentally bought cables I thought would support high refresh rate HDR between a PlayStation 5 and our new TV but it turns out I Did Not Do That.
So I think all I’m saying is that there’s a clear user experience value proposition (ugh word salad) and conflict between:
- The store where you can buy anything
- The store where you can sell anything
Amazon’s model in a way is that this works out great because you do want to buy anything, sellers want some place to sell and Amazon gets to sit in the middle and operate the buy/sell infrastructure and take a (from its point of view) entirely reasonable cut for delivering this marketplace value to society.
But, you know. It’s not the only way.
Another way of looking at this is: what do you mean by everything?
Every product? Every variety? Every customization? Every brand? Every OEM version? Every white-labeled version?
Or even more pithily: too much of everything and anything is bad for you.
Deceptive Design
A quick note that I was fortunate enough to be invited to the Towards Trusted Design workshop, hosted by the Web Foundation and run by Simply Secure and 3x3. Some stuff caught my attention and is in my head now:
- Bounties, a la bug bounties, for discovering and reporting deceptive design (how would they even work? Would they work?)
- “the X-Prize, but for the opposite of deceptive design”
- One-click bug-reporting but for deceptive design, even to the extent of “send a screnshot to $REGULATOR and circle in red with your finger where the bad company hurt you”, because there’s still an asymmetry in reporting, plus I hazard a bet that reporting to national regulators is somewhere on the scale of “fill out a government form”
Okay, that’s it for today!
How are you doing?
Best,
Dan
-
ORNL’s Frontier First to Break the Exaflop Ceiling, Top500.org, May 30, 2022 ↩
-
The Path To Extreme Computing, Erik P. DeBenedictis, Organizer Sandia National Laboratories, Los Alamos Computer Science Institute Symposium 2004 ↩