Comment by Animats
1 year ago
Most of those take a lot more time and money than YC usually offers.
There are some opportunities in "New Defense Technology". Something like a low-cost replacement for the Javelin anti-tank missile based on off the shelf phone camera parts ought to be possible. Of course, once that's out there, every insurgent group will have some.
"Explainable AI" is really important.
"Stablecoin finance" is mostly how to make sure the issuers don't steal the collateral. Maybe the people behind the stablecoin have an explosive collar welded around their neck. If the price drops, it detonates. That might work.
"Applying machine learning to robotics" has potential. Get bin-picking nailed and get acquired by Amazon. Many people have failed at this, but it might be possible now.
"Bring manufacturing back to America". Is it possible to build a cell phone in the US?
"Climate tech" - think automating HVAC and insulation selection, installation, and analysis. Installers suck at this. See previous HVAC article on HN. A phone app where you walk around and through the building with an IR camera is one place to start. Map the duct system. Take manometer readings. Crunch. That's do-able on YC-sized money.
Fun fact about the Javelin bit. I know it was just a throwaway example but something to tickle your brain about it:
Phone camera parts would be overkill. The Javelin sensor isn't nearly that high-resolution, we're talking low triple digits in "pixels". It does however refresh its readings very fast, a necessity given its speed. The old Javelin used active cryo and a filtered IR imager, the new one is passive like the IR camera in some phone attachments. It is stupidly simple in operation: CLU provides the target "signature" and the imager seeks it. After the initial ascent in top-down mode, a stronger signal on one edge of the sensor pushes the control surfaces in the opposite direction until it strikes its target. I'd give the Wikipedia page a read. It contains a surprising amount of information that informs the design and thought behind the missile. Military systems are cool for how robust yet simple they are.
There's a teardown of a Javelin missile guidance system on Youtube! https://www.youtube.com/watch?v=11_5TB0-lNw
I used to work on the Javelin guidance system "seeker head". It's wild seeing him tear it apart. Questions like "what is this for" make me laugh. It's all seen as dead simple, but a lot of engineering went into manufacturing them even as recently as a few years ago. Funny he thinks the sampler chip was expensive, he should look up the cost of the optics!
One thing I’ve been thinking about re: Javelin (and MANPADS)—would it be better to be able to fire them remotely? ie let soldiers put a set of launch tubes in a bush or behind a rock, then use the targeting system from a separate location. That way, the solders’ location isn’t revealed by firing the missile. Better yet if you can strap the tubes to a robot dog.
Another idea: drone AWACS. I mean, a drone with radar to detect other drones (and other aircraft).
The Stugna anti-tank-guided-missile seems to indeed have this feature:
"...that is connected to the firing unit by a cable, allowing it be used at distances up to 50 metres (160 ft) away"
Source: https://en.wikipedia.org/wiki/Skif_(anti-tank_guided_missile...
Doing anything remotely means using radio signals to control the thing, and that becomes the main weakness: radio signals can be jammed. You could try other means of communication, like lasers, though, but those too have their own weaknesses.
5 replies →
"Bring manufacturing back to America". Is it possible to build a cell phone in the US?
I think we should start more basic and work our way up. For example, there isn't a real reason we can't produce all of our domestic iron and steel needs in the USA, but we end up importing a lot right now. Same with aluminum, etc. But this isn't something YC is really going to help with unless they are funding manufacturing and industrial tech that makes it easier/cheaper to set-up and run these types of facilities.
The US currently imports only 17% of its steel, mostly from Canada and Mexico. The US also exports steel, but imports are about 4x exports. So the US steel industry is doing OK.
60% of US steel consumption is now from recycled steel. Nucor became the largest US steel manufacturer by making that work.
It takes (up to) 456.23% import tariffs[0] to achieve that 17%.
So you pay china $1 million for some amount of steel (via vietnam) and then pay the us gov $4.56 million for a total cost of $5.56 million.
It’s amazing that so many steel companies are still underperforming in the USA seemingly in spite the intense protectionism.
0: https://www.latimes.com/business/la-fi-us-vietnam-steel-trad....
1 reply →
Is the percentage of steel imports relevant, when most of the metal consumer products are coming from abroad?
Simple stuff like pots and pans, cutlery, potato mashers. Then industrial parts. Farm equipment. Eventually, cell phone frames and more sophisticated stuff. I think this is what the parent comment is alluding to.
1 reply →
US Steel is now Japanese, btw
3 replies →
You can do advanced electronics manufacturing in North America, it just has tradeoffs - primarily cost and process availability. I work for a company that builds high-end electronics over in Canada; Opinions are my own.
Advanced electronics manufacturing never left the US. The problem is that all the consumer-grade stuff left, and what's left is all really high-end military stuff, and is stupidly expensive. It's great if you're a defense contractor building some state-of-the-art weapons system that really needs the performance offered by those process technologies, but if you want to build a simple prototype for your small business, or you want to build some not-so-cutting-edge electronics in high volume for consumers, it just isn't feasible.
1 reply →
We import a lot but we make a lot. I made a living supervising the manufacturing of the rolls used to roll steel in mill. We weren't exporting even the majority of the product.
Cost. I don't think America should focus on mining raw materials that can be sent elsewhere so cellphones can be made which America will import back at high costs.
> Is it possible to build a cell phone in the US?
Definitely possible, this one is mostly US-built: https://puri.sm/products/librem-5-usa/
It's a good sign that Purism can do this at all, even though it's a boutique/fringe product right now.
Note that it's $1,999+ for mostly made/assembled in USA version, vs. $999 for Purism's made in China version.
And China is more capable at scaling, if the product were ever competing on price rather than principle. So still a ways to go to be competitive at manufacturing.
It's definitely more expensive to produce in the US, but I think it's priced at $2000 because this is literally the only option for people who want/need a US-made phone. I don't think it actually costs them double to produce.
1 reply →
Robotics? The Japanese way?
The is no practical reason why javelin costs the $$$ it costs post r&d which was completed in the early 1990s. The matrix and most other electronics in it are extremely basic and could be obtained off the shelve already like 20 years ago. The concept is already outdated anyway - just use a cheap drone
Some of zhe reason why a javelin costs what it costs:
- small production runs
- obsolete components
- obsolete production technology
- certification requirements
- continued support and design changes to account for the above
- the mandatory defence surcharge
From top of my head.
More or less all of these yes. I always found it ironic how the new Javelin is believed to be cheaper because the components are less mechanical and easier to source. The continued support especially. Military systems can be designed (and warrantied) to last decades if maintained properly - and that costs the big bucks.
24 replies →
I think this is why we're seeing that the type of founders YC usually funds in these industries aren't going through YC and choosing alternative methods of getting started.
Low cost swarm of drones for new defense technology.
We have seen consumer grade DJI drone use in Ukraine-Russia war by both sides.
AI to control a swarm of cheap drones to survey and kill?
The DJI drones are even too costly, the current gen of FPV drones used in the field capable of busting a BMP at 10km distance cost like 400 USD tops.
Longer range systems cost a bit more, but not a whole lot much more.
Improvements that are desirable are generally in terms of range, endurance, sensors and resistance to electronic warfare.
Copying the silent prop design from Zipline would also be neat to reduce the sound signature and give the enemy less time to react…
It’s not like you can hear a drone approaching, while driving APC/BMP, though. For bomber-type quads, sure, it probably can make a difference.
Is this a serious suggestion, or a warning for why there's no place for move-fast-and-kill-things startups in the defense space? No one should be working on this, especially not involving AI in any way.
In Ukraine, hundreds of small teams are working on new drones or other military tech, like radars, camouflage, shelters, demining, etc., but they are low on money.
I personally have some ideas based on my own experience in war, but I cannot make them while in a trench.
> No one should be working on this
Well, given the fact that people are working on it, and many of them probably don't like you so much, it's a good idea for "the people you call friendly" to work on it.
1 reply →
> how to make sure the issuers don't steal the collateral
I think they are looking for a "decentralized" solution where the collateral is held by a smart contract.
> "Bring manufacturing back to America". Is it possible to build a cell phone in the US?
US is offering money, so why not take it?
> I think they are looking for a "decentralized" solution where the collateral is held by a smart contract.
Problem is that unless the asset is virtual, it's going to actually be held by a legal person in the real world, not the smart contract, and it's not ideal to have a stable coin collateralised only by another virtual asset (although makerdao seems to make it work so far).
I think the main innovations here wouldn't be technical but legal, since good solutions to this problem involve the interface between the real world and the blockchain. I remember reading about a legal structure where real assets were held by a trust for the benefit of the owner of an NFT or something, but I'm sure there are other things you could do with the right legal structure, or possibly a central bank or a jurisdiction like Estonia could come up with something that would engender a lot of confidence.
> Maybe the people behind the stablecoin have an explosive collar welded around their neck. If the price drops, it detonates. That might work.
I hate crypto, but I love this idea. We should apply this to a lot of systems.
Make stakeholders of anything accountable. 100% skin in the game.
This is a Chesterton's fence situation - we already know the problems of HEAVY, punitive liability and accountability for everything.
But I do think we're leaning way too far towards the no-accountability side currently, and need to shift a bit further the other way.
(But I don't expect THAT to come out of a VC industry where so many prominent people and parters have track records that generally include a lot of "founded unprofitable company but kept it alive long enough to have a good exit" stories... This world lives on the perception of success, not on long-term responsibility.)
China sometimes does that. Zheng_Xiaoyu, the head of China's equivalent of the Food and Drug Administration, was caught taking bribes to allow tainted drugs to be sold. He was executed in 2007. [1]
[1] https://en.wikipedia.org/wiki/Zheng_Xiaoyu
Wow and only 40 days from sentencing to execution
2 replies →
A pity the USA couldn't do the same to Elizabeth Holmes.
Strongly deregulate nuclear power - on condition that CEO of company operating, designer, manufacturer CEO live with families within 15km of power plant.
(unlikely to work for several reasons, may be stupid idea but looks like something that could work in not-so-different world)
>If the price drops, it detonates. That might work.
I would hate to cross you. That is scary.
What part of a cell phone do you think we can't make?
Do we have the aluminum milling capacity at scale?
Can we manufacture touch screens at scale?
Can we manufacture Li-ion batteries at scale? (Tesla and Panasonic might be able to, with large new investments, but I don't think there's anybody ready to go)
Do we have 3nm fab capacity? (TSMC is planning to build one, but AFAIK not yet)
Do we have the ability to manufacture various sensors at scale? (Some likely yes - ambient light, inertial - some no)
What about image sensors? (Maybe, Omnivision is probably the best candidate, but I don't think they can currently do 48MP. ON Semiconductor is also a good chunk away from that, AFAIK)
I think that's a sufficient number of parts to claim we currently can't make cell phones, as long as you define cell phone as "current gen cell phone". We could probably retool relatively quickly back to at least cell phones, but even that is AFAIK not a current capacity.
Can we _theoretically_ do all that? Sure. But we can't right now, or within short time frames, and we can't without significant investment.
That aluminum milling scaling problem itself is like, an entire category of hard.
CNC machines are hard to scale anything more than linearly. We need to train up hundreds of thousands to become CNC machinists. An entire support industry for machine maintenance, tooling manufacturing (an even harder problem), consumable commodities needs to be similarly scaled in parallel.
4 replies →
The chips. Which are... kind of the most important part, arguably. Hopefully Intel's new fabs will help.
Modems :D