"Those rules of thumb don’t apply to hard tech startups"

January 19, 2017 9:30 AM

How to Build a Hard Tech Startup | CEO Jason Rosenthal deep-dives into the trials and tribulations surrounding the development of Lytro Cinema technology.

Below are the lessons we’ve learned so far about how to build a hard tech startup. Although the examples are specific to Lytro, I believe they apply more generally to companies trying to build anything radically new leveraging cutting edge technology.

Lytro previously

posted by I_Love_Bananas (21 comments total)

Here, you’re often, if not always, tackling a problem that no one on the planet has ever successfully solved encountered.

Keep it real, Lytro.
posted by gauche at 9:44 AM on January 19, 2017 [1 favorite]

This doesn't seem to be so much 'hard tech' as 'any tech that's not web or mobile focused.'
posted by PMdixon at 10:32 AM on January 19, 2017 [3 favorites]

I'm having a lot of fun watching all these tech startups realize they can't disrupt their way out of the age-old problems of manufacturing and distribution.
posted by Admiral Viceroy at 10:49 AM on January 19, 2017 [21 favorites]

I'm having a lot of fun watching all these tech startups realize they can't disrupt their way out of the age-old problems of manufacturing and distribution.

Couldn't have said it better myself. Manufacturing actual widgets is hard, and not the same way software development is hard. When you outsource it to the lowest bidder, you get all the problems associated with production facilities in unregulated working conditions. If you're the kind of person who writes articles like this, you don't read any of the thousands of pages devoted to the centuries-old problems of supply-chain management, and then you end up like this guy.
posted by Mayor West at 11:20 AM on January 19, 2017 [8 favorites]

I'm having a lot of fun watching all these tech startups realize they can't disrupt their way out of the age-old problems of manufacturing and distribution.

I don't think they're trying to disrupt anything, they're just inexperienced and/or naive about manufacturing.

you don't read any of the thousands of pages devoted to the centuries-old problems of supply-chain management

in fairness most of those are about CPG or auto companies so I'll cut electronics hardware startups who think they're going to use all COTS components a little slack.

But like I said, naive.
posted by GuyZero at 11:57 AM on January 19, 2017 [2 favorites]

But like I said, naive.

I'll cop to this naivete myself. I'd worked my whole life in fortune 100 companies and never really given much thought to the supply chain development, management and influence that working in goliath companies brought me as a designer. when I went out into HW startup world I was grossly unprepared for the same challenges he's describing.
posted by Dr. Twist at 12:45 PM on January 19, 2017 [2 favorites]

But it's a pretty cool idea.

It's a cool idea, but it bumps up against the fundamental limitation that it has to trade off resolution. Being able to refocus after the fact is cool, but is it worth going from 40 megapixels to 4?
posted by Pyry at 12:56 PM on January 19, 2017

There's an old saying about logistics and who pays attention to it.
posted by PMdixon at 12:57 PM on January 19, 2017

I actually bought a Lytro Illium a while back, love it: I do blacksmithing, so I set it up on a tripod and use a firebrick lined lightbox to photograph works in progress, not having to fiddle with focus is great -- no autofocus I've tried seems to be able to deal with red-hot steel, with this I can adjust the focus afterwards, and not worry about getting the piece in just the right spot, just slap it down, hit the button, then back into the forge.

Nice to see the company is focused on the long-haul -- I was a bit concerned that when the first couple of camera models didn't exactly woo people, mostly due to the resolution issue, that they might just evaporate, but it looks like they have the right attitude for continued progressive development.
posted by Blackanvil at 1:04 PM on January 19, 2017 [7 favorites]

If your primary concern in creating cinema is deciding what should be in focus, then this product is for you, I guess? Looking forward to hearing a director say, "Just point the camera at the actors and we'll figure it out in post."
posted by Shotgun Shakespeare at 1:13 PM on January 19, 2017 [1 favorite]

I don't really understand the article. #1 says to do proof of concept within some cost function. Why is this a new lesson? Do they have no research engineers on their team?
posted by polymodus at 1:15 PM on January 19, 2017 [1 favorite]

I'm having a lot of fun watching all these tech startups realize they can't disrupt their way out of the age-old problems of manufacturing and distribution.

Couldn't have said it better myself.

Don't sell yourself short; I'm sure you could have said it better. Just imagine this link was written by a MeFite (say, Dr. Twist, who mentioned learning similar lessons). You'd probably express it more along the lines of "these lessons could have been learned more easily in such-and-such ways" rather than something like (paraphrasing) [these failures are funny to me]. So yeah, saying it better is totally doable.
posted by a snickering nuthatch at 1:58 PM on January 19, 2017 [1 favorite]

OK now I see a twist. The problem in articulating a response to this piece is because it's not accurately about lessons in manufacturing logistics for new technologies. It's the nagging feeling that some CEO took it upon himself to appropriate this preexisting discourse (the libertarian narrative where they learn/discover these lessons) which simultaneously serves as ideological marketing and ultimately reinforces the framework of nondisruptive, incremental, etc., industry. And one symptom is that the author was not capable of explicitly reflecting on this.
posted by polymodus at 2:13 PM on January 19, 2017 [2 favorites]

Blackanvil, thanks for explaining why this technology has been useful for you--the scenario where you can't effectively autofocus didn't occur to me.

I'm much more impressed by Metafilter when posts are greeted with more than just lazy snarking. I like this site when we're not just a mob of armchair experts who never show their work.
posted by danny the boy at 2:15 PM on January 19, 2017 [10 favorites]

Lytro had several problems, first and foremost of w - hich was cost. It was eye-wateringly expensive for the quality of images it produced - there are few people that care so much about autofocus that they're willing to pay several hundred dollars more for something that deliver point and shoot quality pics. This was exacerbated by the design choices that made it look like a toy(and thus something cheap) rather than a "serious" camera.

Hippybear, you raise an interesting point, but the reality is that focus is literally the first thing people think about when putting together a shot - why would they need to change it later? It's a bit like having a car. A car has the ability to change destinations any time you're in it - but 95% of the time, you only get in a car because you have a destination already in mind. Occasionally something might come up and you might change destinations mid drive (equivalent of doing a reshoot), but it's not something you do every trip, or even most of the time.
posted by smoke at 3:03 PM on January 19, 2017 [1 favorite]

I got sucked out of the article when it came to the photo of Shenzhen, China...uh, hey editors, that's a picture of the Oriental Pearl Radio and TV Tower in Shanghai, China. Easy mistake, I know.
posted by Existential Dread at 3:08 PM on January 19, 2017 [5 favorites]

Wow. Had never even HEARD of Light Field cameras (is the capitalization correct?) before tonight. Really quite cool.
posted by Samizdata at 7:14 PM on January 19, 2017

Being able to point a camera at a scene and then make adjustments for focal depth and exposure later (along with color grading and other traditional things in post) would 1) speed up filming a movie greatly and 2) would make movies a much more creative medium if the post production tools are available to take advantage of what a light field camera is capable of.

#2 is already happening. Check out the video above point #2 in TFA.

It basically generates a depth buffer for the video, which makes it trivial to isolate scene elements. No greenscreen or matte work necessary! The audio isn't working for me, but it seems they're even able to easily remove crew working behind the actors, as in the wedding example.

This is a very common technique in 3D graphics. Being able to apply it to live video is pretty neat.
posted by neckro23 at 7:37 PM on January 19, 2017

This is kind of curious framing for a company that's been around since 2006. Their initial product was cool but it didn't have much traction. This video camera is arguably a pivot, addressing a different market, and using off-the-shelf sensors (I assume the earlier stuff was custom). So this is Round 2, or even Round 3 for this 'startup'.

It's interesting though to see them go back to the technique used in the founder's PhD thesis, which takes in the data from a grid array of cameras in order to measure the 'light field'. And given that Lytro hasn't exactly taken the world by storm, the massive up-front investment for a custom imaging chip would indeed have not made much sense.
posted by Standard Orange at 9:37 PM on January 19, 2017

Once bandwidth doubles another few times, it would be interesting to see movies (3D or not) where what you're looking at is tracked and brought into focus for you, as if you were looking at it IRL. Right now, directors decide that for you, but there's no particular reason that they have to. (Okay, there is a reason - you'd have to do a lot more cleanup in post to make sure the images were suitable at every depth, which would be very expensive. But money, schmoney.)

When I started the article - a paragraph or two in, with all the talk of doing hard development of brand new technology - I was expecting, I dunno, some kind of basic materials research; the invention of a new kind of image sensor, say. Instead, it turned into an article about sourcing parts from here and there and putting them together in a new way. It was reminiscent of the modern web development that the author refers to - grab a jQuery library from here, a little D3.js from there, a little Ruby on Rails, and mix them all together into something new and revolutionary. But with hardware.

I suppose I was naive.
posted by clawsoon at 8:25 AM on January 20, 2017

Is it just me who's worried about AI these days. Can you imagine technology in 50 years time...

We are entering a new age for sure...
posted by metafiltermonkey at 11:27 AM on January 22, 2017