Ask HN: What startups are working on hard, technically challenging problems? | Hacker News Reader

Ask HN: What startups are working on hard, technically challenging problems?

36 pointsmtae23 comments

I made the same post 4 years ago and there were some great responses. It's pretty outdated by now though so here's round two.

Link to previous thread: https://news.ycombinator.com/item?id=9242217

In the article, "Silicon Valley's Youth Problem" [1] the author mentions Meraki (now Cisco-Meraki) as an example of a startup working on advances in technology rather than the latest web app. Do you know of other companies that fit the description?

Follow up: Are they (you?) hiring?

[1] http://www.nytimes.com/2014/03/16/magazine/silicon-valleys-youth-problem.html

View on HN

I have this old dream of building a great verification / formal methods as a service company with the right mix of axiomatic specifications, data and flow analysis, abstract interpretation, model checking, type systems and theorem proving.

And with time extract project methods into tools and frameworks, and evolve them. A bit like 37signals did by extracting Ruby on Rails from Basecamp.

A source of inspiration is bridge engineering from the mid-to-late 1800s, when things became very rigorous thanks to applied mechanics and structure theory.

Galois and SRI were doing some work in the area, but it has not become as big or central to their business as I would have expected.

https://en.wikipedia.org/wiki/Semmle has been doing some interesting dataflow and static analysis for some time.

Latent Sciences: https://latentsci.com Building legit AI solutions (not deep learning (wat?)) to fight the most challenging diseases with broad impact.

Yes hiring: https://angel.co/company/latent-sciences/jobs

And to recommend a company I don't work for: https://tlon.io/ It's not yet-another-crypto, it's not faux-AI, it's real technology for the sake of building something new.

I tried to understand what is tlon’s product (urbit), and failed. What is it?

What is Urbit? https://www.youtube.com/watch?v=M04AKTCDavc

Don't know if I'd call it a product per se.

Yeah, they lost me at the first mention of a “personal server”. Let’s pretend I’m a farmer in Kansas. Why do I need it? What can I do with it?

You don't. And probably not much.

I work at https://limejump.com and we have a platform which manages energy assets such as wind farms, batteries and diesel engines and helps their owners make money by trading on the energy market.

It's technically challenging because doing so requires a team of data scientists forecasting the performance of the assets using weather data etc, an IoT team working on devices which send real time data from the assets, a software engineering team building the software behind all of this and an electrical engineering team that works on the assets.

The description above is a gross simplification and I've only been here a few months so I'm still getting to grips with things but it's definitely full of technical challenges.

I have seen only a very small number of startups genuinely face very hard engineering problems like what Corellium is doing. I will elaborate.

In a 9 year development span, they have fully virtualized a real iOS device (Yes, it is not a simulator) to the point that it is running on a rack of ARM64 hypervisors on bare-metal boards from the bootloader to the home-screen. They even managed to virtualize the secure-element and access the console from there.

A task like that is something that is very rare in startups these days and to get it fully working like this is a herculean task, because the interals are closed-source and undocumented and was certainly reverse-engineered. Corellium is also at the point at which Google is using it to cross-compile Golang builds for iOS and Android. That is how good it is.

I think this startup changes eveything in the mobile industry.

Why do you need a virtual iOS device? Honest question.

Perhaps you should ask every company testing apps on real iOS devices and the security researchers who need to automate finding bugs in iOS apps and in iOS itself. The iOS simulator is severely limited and is controlled by Apple so the need for a service that runs a full unrestricted iOS in the cloud is in demand and Corellium solves this.

Apple has maintained control of the mobile industry for a very long time and Corellium removes the need to buy another locked-down iDevice due to planned obsolescence to test your app. There is a reason why Apple tries to shutdown anyone who attempts to virutalize or emulate iOS/tvOS/watchOS and their apps.

Why does Apple limit their iOS simulator? Do you need iOS simulator for development of iOS apps?

The simulator only runs intel versions of iOS apps and cannot run true iOS apps which are arm64 binaries. This is Google's use case for testing Chrome and Golang on a truly virualized iOS device capable of running these executables. The iOS simulator cannot do this.

For testing smaller applications, the simulator is sufficient for that purpose, but for more complex applications almost always requires a real device. Also some features like in-app purchases are unavailable in the simulator, making it harder to test such features.

The real reasoning behind why Apple uses a limited version of the simulator is unknown to everyone, but these limitations have forced developers to purchase real iDevices just to test certain features either when Apple ends support for a device or an iDevice doesn't support a certain feature.

Corellium takes this further and fully virtualizes the iDevice on any version and it has the options to configure the state of the iDevice and allows automation and tests to run on the device. You may think of it like what Docker did to web apps is what Corellium has done for iOS and Android apps.

Can I use this simulator to make phone calls?

Seems like Rigetti is working on some pretty cutting edge stuff: https://www.rigetti.com/

OpenAI also comes to mind: https://openai.com/

Relevant: https://news.ycombinator.com/item?id=18958321

We were presented with an interesting challenge from a potential lead for our edtech platform where they want to be able to run an online course platform in countries where internet connection is a luxury. So it has account for bad/low/unstable internet connection but still provide a way to learn online. For example, when we build applications like those, we don't think about the internet connection necessary since we assume it has to exist. But offline sync etc for this learning platform will be key. This is not an entire startup by itself but part of our current business and a fascinating challenge to solve.

https://archive.org/details/Microsoft_Encarta_The_Complete_M...

Singapore has some really cool ones tackling problems in Quantum Mechanics, Computing, Space Tech and Biotech.

An example is Atomionics http://www.atomionics.com/ (Disclaimer: a friend's company).

Plus, Entrepreneur First here are generally putting in good ones. List of companies here - https://www.joinef.com/companies/location/singapore/

At GrAI Matter Labs we are working on next gen non-von Neumann brain inspired computing assuring post-Moore performance scalability. This technolog is only moving now from research to production and has been labeled as transformational by Gartner. Our technology fuses neuro science and computer science in one architecture that is both, trainable and fully programmable.

And yes, we are hiring. https://www.graimatterlabs.ai

"labeled as transformational by Gartner"

According to wikipedia, there are 64 magic quadrants with roughly 10-25 companies per quadrant. This type of marketing is more of a detractor than anything else, especially on HN.

It's much more helpful to provide technical / product related information to substantiate your claim. If you are working on a hard problem, describe it. The more detail, the better.

The hard problem is to provide processor performance scalability in a post-Moore world. We are tackling this by event based, near-memory, sparse computing. Traits that can be just as well attributed to the brain. The challenge here is that a whole new set of algorithms is required. A good illustration for that are DVS (Dynamic Vision Sensors). They produce a continuous stream of asynchronous events, just like our eyes, and algorithms like object detection have to work on this stream of events rather than a simple sequence of frames. The advantages are ultra-low latency, high energy efficient computing. Only what needs to be computed is computed, asynchronously. Since DVS is not a very widespread technology (yet) we do frame-to-event conversion to tap into the energy efficiency of sparse, near-memory compute. All of this processing can be done on the same architecture, since it can be explicitly programmed to do e.g. frame-to-event conversion, and trained to do inference on objects in the resulting stream.

Nervos Network is: https://www.nervos.org

Yes, we're hiring. Check out Angel list: https://angel.co/company/nervos-1/jobs

Reply on news.ycombinator.com