Hacker News new | past | comments | ask | show | jobs | submit login
Startups from Y Combinator’s S19 Demo Day 1 (techcrunch.com)
169 points by bruceb 64 days ago | hide | past | web | favorite | 140 comments

> Epic Aerospace: Epic is manufacturing inexpensive space tugs to deliver satellites into geostationary orbit. The 21-year-old founder has been building rockets since he was 16, and is now managing a team of seven aerospace engineers with Epic Aerospace. The founder describes propulsion as one of the biggest problems for satellite companies, in that it can take up to two years to qualify new satellite systems and can cost up to $30 million. The problem they’re solving is moving satellites from low Earth orbit directly into geostationary orbit. Epic’s tug is half the cost of the competition and is reusable. They’re currently working with Satellogic, and chasing what the founder says is a $3.1 billion geostationary insertion market.

So, uh... what's the big propulsion breakthrough? They have no website (well, they have this: http://epic-aerospace.com/) and I can't find any search results. They don't list the founder's name, and the only person I can find or other mention other than the ycombinator article is this: https://github.com/OmarBessa

I agree it's a big market, and it costs a lot more than $30 million to prove out a new propulsion system, especially one that can repeatedly take satellites from LEO to GEO.

If it's some reusable tug, I'm guessing it's able to be refueled in orbit? That'd be pretty epic indeed, because I'm not sure of anyone doing that these days. Though transferring fuel in zero g is a real interesting problem.

Probably is no breakthrough, other than solving a common problem with a reusable approach.

It is wasteful to have every satellite carry the weight of its own propellant and engines.

Having a single device that can deliver quite a few satellites to their orbit through clever planning and efficient fuel use could save significantly on weight, design, and launch, costs and complexity.

Well delivering multiple satellites could be a thing, it's still basically the same cost by mass, you're just splitting the monetary cost up among different customers. If you are delivering multiple GEO satellites, you also have to have them be the right place in their orbit, over their target, since they are stationary with respect to the earth (which is why you don't have to move your satellite TV dish). This means all your satellites for that bunch have to basically be near each other.

That also doesn't feel like it is reusable given this definition. I'd assume reusable means you can use it now, and then use it next month on a completely different mission. So let's explore that.

GEO satellites are also generally a lot bigger than your typical cubesat / nanosats that are all the rage in the bulk launcher configurations, because you need a big dish, and a lot of power to transmit and receive signals.

> It is wasteful to have every satellite carry the weight of its own propellant and engines.

Even if a satellite was delivered to GEO without using an engine, satellites need engines to do stationkeeping and keep their orbit, which is the critical factor for the operational lifetime of a satellite.

It's more wasteful to have to have your orbiting tug do an orbital rendezvous with a satellite in LEO (takes fuel to rendezvous), go to GEO (the normal cost, but more expensive since you have to carry the fuel for the return trip), then come back from GEO to LEO (takes fuel).

Following this delta-v map, it looks like it takes about 4000m/s of delta-v to go from LEO to GEO, then the same the other way back down.


I don't know of any clever planning or efficient fuel use that can get over the raw amount of delta-v you'd need. Fuel use isn't really efficient, it's the engine that has to efficiently use the fuel, which is why I think there must be some breakthrough here.

Fuel also has to eventually be delivered up to the tug for refuel, so really it seems like all the mass savings is the dry mass of the 3rd stage.

I totally agree. You still need to get the fuel to space regardless of who is doing the propulsion work. Let's say that your tug is already on-orbit, how is the fuel going to be brought up? Would you not need to launch another spacecraft to bring you the fuel? How much would that spacecraft weigh?

Besides this, they would have to develop on-orbit refueling technology, as well as force the other spacecraft to use a bus that allows for rendez-vous and docking/proximity ops.

And also, how big is the market for satellite startups that are heading to GEO orbit? For now, it seems like it is just Astranis and GapSat.

We will be able to transport satellites up to 5000 lb in mass directly to GEO, which should be most of the market.

Almost every large GEO player is developing new, smaller satellites. We will also modify existing interfaces to allow for docking. Rendezvous and prox-ops is all on our side.

Who is "we"? Care to identify yourself? What existing interface are you talking about? Which GEO manufacturer is making 5000 lb spacecraft?

I can take a wild guess who "we" is, since Epic's young founder is Ignacio B. Montero and the nick is "ibmont".

"you got me"

That you don't know of it doesn't mean it doesnt exist :)

That's why I said "I don't know of any." I would love to know more if you know anything about what they're doing. As the only reference and information I could find about this company was literally this one paragraph (and I tried), it's not enough to say it won't work, but I think it's the company's job to explain how they will make good on their claims.

Are you involved with this company?

>it can take up to two years to qualify new satellite systems and can cost up to $30 million

That's the breakthrough. A sat manufacturer does not have to design, build and qualify their own propulsion system, but can pick this off-the-shelf part from Epic Aerospace.

The equivalent in electronics would be switched power supply modules in the 80s. When they were new, a PC/TV/hifi/other high tech product design team would design their own switched power supply, spending engineering time on that. Ten years later, they would just pick one from a supplier catalogue, and spend a lot less time on just integrating it.

> but can pick this off-the-shelf part from Epic Aerospace

And there are no existing off-the-shelf suppliers for satelite propulsion systems? That sounds unlikely.

No these systems tend to be custom due to cost, reliability and size constraints. More so for GEO satellites which tend to be the most expensive.

That's what I was thinking too. Those are all for Cubesats tough, none of which have gone into GEO so far.[0]

[0]: https://space.stackexchange.com/a/22721

I didn't research every spacecraft but if you look at GTOSat, its intended orbit is GTO, so it won't really need the services of Epic since a rocket can place them there directly.

And if you look at the spacecraft, it doesn't seem to have any kind of meaningful propulsion system. https://space.skyrocket.de/doc_sdat/gtosat.htm

I'm sure the other spacecrafts have a similar design concept.

Operating in GTO is terrible for electronics.

And also, we could do GTO 2X cheaper than large rockets for even their nominal payload size. For smaller payloads, it's more like 10X.

So is your idea to have spacecraft go to LEO first instead of direct GTO? Is this an efficient use of fuel?

It's not a detour, you have to pass LEO to get to GTO. Instead of the rocket pushing it to GTO the tug will take over. So the rocket has more payload due to only having to go to LEO.

Props to the founder for his foresight choosing a name that will perfectly position them for an acquisition when Tim Sweeney decides he's ready to go head to head with Musk and Bezos.

Space tugs have been a topic of research for quite a while in the space community. A number of companies targeting the in-space space transport market over the last few years, e.g., [1][2]. There's also a recent YC alum targeting this [3].

Propulsion is itself a really hot topic in the NewSpace market the moment, ranging from tiny thrusters for precision attitude control of smallsats, to large systems for orbital transport. We recently published a review of smallsat thrusters, highlighting the rich spectrum of solutions being brought to market [4] (Disclaimer: I'm one of the co-founders at satsearch).

There's a lot of systems innovation that enabling fuel-efficient solutions to be developed. Particularly, there's a lot of progress being made on developing high-efficient electric propulsion. One of the trending areas is water-based propulsion, given that it's clean, non-toxic, storable, and also abundantly available in the Solar System [5].

With companies also look at building fuel depots to refuel spacecraft on orbit [6], there's a wide array of new solutions that are going to emerge for repeat use of tugs.

The traditional GEO market went through a massive dip last year [7], but there's renewed interest with the real prospect of small GEO missions materializing in the next 24 months. Another YC alum is targeting this market [8].

So in short, this is a really dynamic sector, going through a lot of changes at the moment. I'm eager to learn more about Epic, to see where they're carving a niche for themselves. Also, interesting to see YC's continuity in funding hardtech space startups.

[1] https://spacenews.com/airbus-to-challenge-ssl-orbital-atk-wi...

[2] https://www.atomosspace.com/

[3] https://momentus.space/rides/

[4] https://blog.satsearch.co/2019-07-10-cubesat-thrusters-an-ov...

[5] https://www.thespaceresource.com/news/2019/2/propelling-dema...

[6] https://www.orbitfab.space/

[7] https://spacenews.com/amid-geo-downturn-launch-operators-loo...

[8] https://www.astranis.com/

>Juno College of Technology: JCT is creating the technical university of the future. The startup operates a coding bootcamp, expected to do $3 million in revenue by the end of 2019. Similar to Lamda School, they offer income-share agreements, but “the similarities stops there,” explained the founder. Juno says it places 87% of founders who complete their nine-week long program.

How is it different from Lambda School? It’s not apparent from their website other than it’s based in Canada and they have a physical campus in Toronto.


This is a re-branding of HackerYou, they've been around for a while.

Lambda School is 6 months vs. 9 weeks with Juno.

Wait, so people actually pay 17% of their income over two years for a 9 week course? What on earth is the upside of this over picking up a few books/online courses and attending a couple of networking events?

One possible upside: Not all jobs are advertised and maybe Juno/Lambda have worked up agreements with recruiters.

Also,think about it from a recruiter point of view: Would you as recruiter rather pore through hundreds of resumes from people who have picked up courses online (and risk getting beat up by interviewers when a few of these don't perform well) or have a pre-screened/evaluated candidates? Worst case: you pass the buck on to Juno/Lambda. There is some incentive for a recruiter to look at Juno/Lambda candidates,no?

I get what you're saying but from an employee and personal development point of view, even the most intense 9 week course sounds guaranteed to lack the necessary substance and depth I'd want for a career start / change (which is what they're marketing this for, right?). The same argument would make me just as wary from a hiring standpoint, don't see much of a difference of qualification between 9 weeks of bootcamp versus a selection of relevant online courses there tbh. Worst case in my mind would be a $startupcollege candidate that was just trained on passing through interview processes, that's not sustainable for recruiter/employer relationship in the long term either.

That being said, my experience with letting recruiters filter candidates haven't been the best in the past and my exposure to bootcamp grads is limited since that's just not much of a thing in the German market overall (I guess because there's no real need since traditional university/apprenticeship education paths are far more affordable around here). Is hiring these candidates common in the US outside of the usual startup circles?

> Is hiring these candidates common in the US outside of the usual startup circles?

I’m afraid I don’t know :-). I’m from India this isn’t common here. Yet.

> What on earth is the upside

For the attendants or for Juno? ;)

Is this the bootcamp version of the famous Dropbox HN quote?

I only partially agree with you. I also prefer learning with books and online courses but that's not everybody's case.

Some people benefit much more from a structured process where they can ask questions easily, be tested much more often and gauge their progress over time.

The example of online courses was merely because I can't think of another equivalent to this crash course approach to education.

While that might be in Dropbox territory as well, I don't see anything other than monetary reasons for why somebody would prefer a 9 week course. Availability/geography can't be it for purely US based programs. Traditional university offerings have not only as much structure as you'd like but also give you knowledge in depth if you choose a good set of courses. All the while providing a pretty customisable, and mostly sane, pace through programs that gives you as many challenges or as much freedom to explore as you want. Not to say that universities shouldn't improve but I'm just quite bazzled as to why the market seems to think these bootcamps are such a great idea.

I think it was Make school who said they did 12 weeks and that wasn't enough so they expanded to a 2-4 year bachelor program

for my kids’ university I’m assuming 8+ years...and that’s if they don’t follow Mom’s path through multiple PhDs!

All so they can program the robot caregivers that will swaddle me someday.

I'm almost going to graduate in HS and hopefully can either get picked up by YC or theil instead of going to college. Given the high chance I do go to college, I'll just be doing whatever I'd be doing at YC anyways.

I also thought about doing academia for a long time, but realized its just a sticker mark on your resume. Given that I graduate from undergrad, I'll hopefully get a good sticker doing masters.

> The startup operates a coding bootcamp, expected to do $3 million in revenue by the end of 2019. Similar to Lamda School, they offer income-share agreements, but “the similarities stops there,

I'm interested in learning how this can be VC scale. From what I understand, it is a very people-intensive business (look at the student:teacher ratio for one of their courses)[1]. How is it different from,say,a consulting firm in terms of scalability?

[1] https://junocollege.com/course/web-development

Short answer is to build a business that’s product-led and incentive-aligned. As long as operationally they’re trying to use technology to reduce friction at every step (student experience, instructor experience, and curriculum) they’ll be able to scale better than an organization that bills for time (and is incentivized to use always consume and bill for more time!)

Can't some people-intensive businesses scale to huge size? Look at Walmart

I went through the list and these are the ones that intrigued me.

lumineye: I think this sort of technology can be very useful and I'm interested to see how it is used in the real world.

safely deposit: I like the simplicity, instant value to users, and the way it transitions physical documents to virtual ones. I think the price is a bit steep though.

puzzl: This seems like it erases a huge headache for its customers. I think about this one as provisioning people like VMs to be a physical presence. Why hire when you can rent on demand...

million marker: There are many people who are caring more and more about toxins in their environment, and this plays directly to their growing paranoia.

nomad: I personally don't use ride sharing, but this one seems like it could be competitive. It doesn't seem like there's much difference between the current options.

earth-ai: Of course! If they have a good solution then they may be able to find a lot of resources that we haven't fully exploited yet. This is the big one I think.

re: Earth AI

Discovery of unexploited mineral deposits is relatively easy already, this is not a big driver of commercial viability or mining economics. That will be a difficult part of the value chain in which to make money, the nominal value of extractable mineral is very weakly connected to market value of a deposit. A rich concentrate of valuable metals does not imply economic value in practice. Extraction economics, on the other hand, is everything and can vary widely. Most mining companies already have a huge back catalog of unexploited deposits that they do not count against reserves (and most mining companies are incentivized to lowball reserve estimates in any case).

> Nomad: In a two-month illegal trial period, the company facilitated 5,700 rides at Indiana University before the startup had to shut down, but they say they’re legal now and ready to try new markets.

Yes, this is a good look for a legitimate business.

Easier to ask for forgiveness than permission

safely deposit: Does seem convenient, but I'm not sure I could get over my privacy concerns.

I’ve been wondering recently - why there are almost no security startups in the YC? Cybersecurity is one of the biggest problems in the 21st century, the market opportunity is huge and solutions require to solve hard engineering problems, yet there are no cybersecurity companies funded.

Does anyone have the same feeling? Why it would be this way?

I'm currently in YC startupschool with a security startup, here's what I think is the problem: security doesn't scale easily on the marketing side. For most businesses security equals costs, and no quantifiable savings.

Most business don't have budget allocated at all for security. Big corps are the exception, but then you'll have the trust issue: they are unlikely to do business with startups when it comes to security.

I have also found that it's hard to find security experts who are willing to work for a startup. Given the salary options that security techs can get at big-corp/government/military, it's hard to be able to compete.

Yes - agreed. I'm working on a product that started out as a a security SaaS, but has pivoted (using essentially the same technology) to an adtech product, because it is much easier to sell an increase in return on advertising spend than it is to sell a security product.

Most security companies seem to fall in one of two categories: - consulting - appliances (aka blinky boxes)

Consulting seldom requires venture capital and cannot scale at the level VCs expect.

Blinky boxes can, but customers are slowly catching on to the fact that adding another blinky box to the network is really not improving their resilience all that much.

On top of that, it's not always easy to identify the real economic buyer in an organisation, and e.g. a CISO might not have the authority to decide on installing new appliances but needs to collaborate with e.g. IT. It all makes for convoluted and longer sales cycles.

While its not in YC - there are a few SAAS type security startups aimed at SME's - such as https://intruder.io/ who I am a very happy customer of. This to me looks to be high growth.

There are dozens of these SaaS vulnerability scanning companies, with very little differentiation between them - they basically all run Nessus. It's difficult to see how they could make serious money, certainly not enough to interest VCs.

The only way is to make them free and build a marketplace of accredited and vetted pen testers who you charge to be on the platform. Still how the hell you vet these people is a huge issue.

I think that's what bug bounty companies are doing by changing the model from paying for time spent to paying for (accepted) findings.

Personally I don't think any serious pentester will spend considerable time in this model though.

That said, I also think that one of the main challenges for buyers of pentest services these days is to evaluate the quality of a report or of the work done. Thoughts on how to improve that are most welcome.

Mostly they are cookie cutter bullshit, right.

https://cobalt.io/ works more or less like that

Yes, I think the other problem with this approach is the terrifying consequences if the blinky box get's hacked.

I think the trick would be - somehow allow the blinky box to be secure (how do you even do updates if the networks you are "protecting" are not on the internet), this is difficult. And then giving the boxes away for free with a limited amount of stuff they can do without pen tester involvement.

Still is massively problematic.

Some startups don't fall in those categories, for instance https://www.sqreen.com which recently raised $14M from Greylock.

And they wet through YC :D

My opinion; it has nothing to do with YC... in my experience (worked building software at a security firm) people who are good a breaking things are generally bad at building them.

Check out https://cylonlab.com/

From participating in that, my experience was that it’s really hard to disrupt in cybersecurity because businesses are skeptical about using young companies with no reputation and a beta product.

So it’s hard to start showing revenue or traction to get funded. The hunt for revenue then also forces you to focus on things like phishing protection instead of discovering zero days.

Sqreen (YC W18) is one counterexample, but they came into YC at a later stage of maturity than your average YC company.

As to why there aren't many others, that's a great question. It's a hard space to get started in, since security is mission critical for essentially everyone, so that likely doesn't help matters. There's a lot more that needs to go into a viable MVP in cybersecurity than in other areas as well, since there's much less room for error and jank.

Those reasons likely contribute, but none of them are hard blockers.

Security is like having an insurance. So now is the question: pay for insurance now or pay after crash? Or hope it never happens? Many many people underestimate this need and end up not paying. I can guess from experience with secure Bluetooth headset startup, that the need isn’t there yet. Maybe I’ll try with Bluetooth 5 hoping for more demand in couple years.

You also might not even know that you crashed in the wall 3 months ago.

We exist, but we're just not in the right country :)

From my experience working with a number of companies, the challenge seems to just come down to a lack of people in the industry.

Are no cyber security companies (1) being founded, (2) applying to YC, or (3) being accepted by YC?

Scared to comment in case I end up like the guy that slagged off DropBox.

Maybe I should do a separate comment for each startup saying "Great Idea! Guaranteed success!" so I can refer to it in five year's time and look really prescient.

Only one crypto mention, and it's a company selling banks tech to detect crypto fraud.

TRM Labs: Banks are required to trace the source of their customers’ money. TRM helps banks identify and trace cryptocurrency fraud. They charge $20K per user seat. Though they couldn’t say the name, TRM says they recently signed a top-five global bank as a customer.

Mighty is interesting, though I wonder if they do anything fancy to avoid being a trivial MIM proxy for your traffic? E.g. do they somehow prevent themselves from logging onto your VM / VPN?

How does http://www.nomadmoments.com/ survive the massive war chest and tech advantage of Uber/lyft?

The $25 a month fee seems easy to get drivers to pay as they can pay that to uber/lyft everyday in commissions.

I suspect the subscription will not stay at that rate for long.

Lyft and Uber are public now, so they have a bit less latitude to engage in an even more aggressive price war, at least without making some painful cuts. They’re already under pressure to demonstrate that they can be long term going concerns. So, I think Nomad is coming out at a perfect time.

Also, it’s much cheaper to jump in and piggyback on the existing supply of drivers and riders than it was to create that market. It’s one more app to run/check, and many drivers already run two, and many riders check two. If Nomad runs super lean and they offer drivers and riders more income and cheaper rides, respectively, without singing the experience too much, I think they have a shot.

In any case, I’m rooting for them, it seems like it’ll put some more power in the hands of the drivers, away from Uber.

One ride at a time.

The "tech advantage" is not insurmountable. These are, at the core, just apps that let users hail a ride and pay. Uber-Lyft competition has already demonstrated that ride-sharing doesn't come with an automatic moat. Both drivers and passengers are willing to use multiple apps and both respond to prices. Hence price wars.

The big problems are network effect chicken-egg problems. You need enough drivers and passengers simultaneously. But, this only needs to be solved in one locale at a time.

Uber & Lyft put all their efforts into scale, assuming that winner-take-most dynamics will fix the economics eventually. They didn't.

It's not a terrible idea to start a competitor building a low cost/fee model from day one. Flat fees for drivers might be interesting because once the fee is paid drivers have an incentive to get their money's worth.

It's not an easy win to go after, but not impossible either. Ride sharing is not an airtight market.

Sorry, this is kind of off topic, but why the hell is such a simple website 7MB? I instinctively dont trust companies that cant even deploy a simple website. The massive background image doesnt even look right on mobile, its scaled to fill.

They're trying to get to market.

The team is probably the founder and his little brother and the cat helping out when it can although its paws slip off the keys on the keyboard.

They have more important things to do although I do concede it's a pretty big glitch.

They'll get around to making it awesome.

This is really not the important thing here.

The hard part is going to market and carving out a niche for yourself (not to mention the endless customer validation and following synthesis), especially in a crowded and rapidly commoditising market like this.

Spending a long time on your site in the early stages is a completely irresponsible and wasteful use of time, when you can be developing or outside validating.

It's way worse than that.

It's served 25.9MB to me.

They are using ginormous background images, and loading multiples.

8000x4000 jpgs. 32 fuckign megapixels. For a background image.

I had it even worse:


32.6 MB

And I have to agree with the OP, simple things like this instantly drive away my trust that they know what they're doing.

>>> simple things like this instantly drive away my trust that they know what they're doing

It inspires me with confidence that they are focused only on things that get them launched.

That works great until it doesn’t?

What else have they ignored that is going to bite them in the ass?

Yeah okay, forget first impressions. It's such a basic thing, who knows what else they've screwed up.

Bounce rates and load times are important if you want people to try the product.

Sorry about this, but we haven't figured out how to scale images to different dimensions, so get in a car with one of our drivers who we didn't background check and is probably drunk.

"Do things that don't scale" Paul Graham

> things that don't scale > drunk driving

You're not wrong.

funny you say that, it was behaving a bit weird on FF desktop for me.

Not secure either…

Yeah I really do not see the benefit of a fixed fee as a starting point competing against a %. Doesn't make sense at all.

Why switch to a new service and be 100% likely to pay something, yet not 100% likely to get paid yourself at all?

Would rather start out with a 0e/month and get it up to, say, 50$ (or more) based on a rolling basis based on your revenue per month.

The app looks like a thousand other ride-hailing attempts as well. What the hell is the secret sauce here?

Congrats to all the startups making it to the end! This was quite a diverse array of verticals: from rare earth mining to mail-in sperm tests ;)

Green Tiger. The "Robinhood of India". We hear a lot about how private wealth management is an untapped goldmine in Asia. To the tune of $100T+. And I'm wondering if the barriers aren't more customer development oriented than a lack of actual trustworthy services?

How Asia Can Protect It's Crazy Riches


$100T is an ambitious number. Whe World GDP is about $80T.

$100T? Source for that?

From a privacy and security standpoint, can some help make Mighty make sense? I just don’t see it. Yes, Chrome does hog my RAM and CPU, but my browsing experience is simply not bad enough to just throw in the towel and give in to a MITM

I mean, I agree. It seems like a backwards-facing technology. Smartphones and netbooks are getting better and better processing power, that power should be harnessed, not just ignored.

Besides the obvious privacy and security issues, I can see how you could get customers in the short term with savvy sales and marketing, but trends in development and tech don't seem to support this being a sensical approach for the future.

I mean, one of the examples given is Slack. Slack is already web-based, if you run it as a browser tab, you don't use much extra CPU/memory. But many run it in the Electron "desktop" app. Why? Because it's more integrated with the operating system and more like a first-class citizen app.

That will be solved now with PWAs and the new Web APIs being worked on, it's just a matter of time imo. When all these electron-based apps can just re-use the already running Chrome browser's rendering engine, and still look like a standalone app, who cares how much CPU Chrome uses?

It'll be basically the only UI app running on the whole machine.

Nothing targeting climate change? Hopefully day 2...

Edit: mea culpa, Holy Grail is a play here of sorts.

Also, Wren previously on Launch HN https://news.ycombinator.com/item?id=20461577

Also Wren, which did a Launch HN: https://news.ycombinator.com/item?id=20461577

From a comment there about planting trees or preventing deforestation...isn’t the problem now that climate extremes will make it harder to expand forests with droughts and fires? Seems like it may have worked 50 years ago but too late now.

I think we need to start investing in ideas that remediate actively and aggressively. Like “oh my god there’s a killer asteroid about to hit us” level of motivation.

I mean the planet will do just fine without us, but if we want to stay the dominant species (and not jellyfish), we had best do as inventors/investors what politicians and governments cannot.


* Make totally off the grid solar easier and cheaper than PG&E (remarkably hard to do today)

* zero petroleum scalable ag

* zero petroleum mass consumer packaging

* recycling that actually recycles

* economically profitable global scale carbon capture that can work in deserts and fires and extreme cold


Edit: Because this: https://news.ycombinator.com/item?id=20745486

Does anyone here feel the need for PopSQL?

I am asking this because I was building on a similar idea a couple years back as a side project. I lost track but got curious seeing PopSQL.

I got quite excited by that entry. My immediate reaction was 'oh, that's really cool'. But after thinking a little further, I can't think of an audience that both uses SQL frequently and also couldn't use a git repo. What kind of user demographic did you have in mind?

My target audience was basically every non-programmer in a team who wants to see/track any kind of data for their own product.

The SQL building would be for programmers to setup say a few initial queries. But anyone can change (there would be a changelog) and share new things to track.

Reportedly you have a lot of customers. How did you find them? Any tips to share? Thanks :-)

Thanks; I've yet to meet any sql-wielding non-programmers in the wild.

Perhaps it's worth releasing after all?

Just an anecdote, but my dad learned SQL 10 years before retirement. It helped him do his work, which was not software development, without burdening the professional programmers.

Of course, there are plenty of free and slick GUIs offering the basics of Git. I'd like to see how this is better. Maybe customer service?

SQL queries for non production uses feel more like “quick documents” and having to interact with git every time you want to share would be a pain. (Imagine telling people to share google docs via a git repo). And I say this as a coder who loves git. So many companies will have sql in their wiki docs eg in one note or even in slack for ease of sharing.

> an audience that both uses SQL frequently and also couldn't use a git repo.

I think you might be surprised. SQL is used a lot in businesses by non-programmers. It's embedded in all sorts of enterprise software, including SME stuff like Salesforce.

"Can't" is a strong word. Maybe they could use git, but most don't.

The Dropbox kind of demographic?

Indeed there's a need... We were putting common SQL queries in the code repository for later copy-pasting in the past, but we stopped doing that when we discovered Metabase (https://metabase.com/). I believe it will be relatively hard for PopSQL to compete with it, since Metabase is light ears ahead as of now.

We use BigQuery where I work, it has a "saved queries" feature. I suppose this could be like a poor man's BigQuery?

There is also a similar open source alternative: Poli (https://github.com/shzlw/poli) Disclaimer: I work on Poli.

There’s also Mode Analytics ( https://mode.com ).

Being able to share queries was one of the more attractive things about Looker in the beginning.

Could somebody explain the thinking behind startups like yummy? This seems to be quite common in the last few years. Do these "robot barista" type installations really take in significantly more revenue than a traditional vending machine that justifies this level of engineering / cost?


Looks like Cafe X (https://cafexapp.com/)

Seems like a gimmick to me. But who knows, maybe their sales are great? Also the real question isn't the upfront (or potentially distributed) engineering cost - it's real estate. Can the robot justify the larger footprint of a typical coffee vending machine by increased sales? My guess is they're targeting the place between vending machines and Starbucks, which seems like a pretty big void in the coffee market. I still don't understand why a robot arm would do anything other than add to the novelty.

Apparently the YC interview days are quite exhausting for the interviewers : because of this in every YC batch there is at least one coffee startup.

It is subliminal stimuli.

>Kern Systems: This startup wants to store information in DNA.

Just imagine the complicated algorithms necessary to prevent literal virus injection attacks.

Some surprisingly interesting companies this round, including quite a few leveraging hardware and robotics in cool ways. That being said, two stuck out to me:

I can't see Lumineye as something other than military and police state tech pretending it's going to help first responders more than it will help special operations units kill people and police/FBI violate privacy even further. This should be disturbing, not cool.

I like what Well Principled is trying to do, but having worked in the consulting space for a long time, the disruptable segment of the management consulting market is not that big. A lot of clients don't hire McKinsey for actual insights or results, they hire for the brand name, someone to point the finger at when things go wrong, and as trusted outsider confirmation of what executives wanted to do in the first place. I have a feeling this will pivot into the financial engineering segment of the market pretty quickly or get bought out by one of the big 4.

Regarding Lumineye, there is a market for elderly people at home, which are already monitored by bracelets or some other devices they are suppose to wear, that they usually forget, to detect fall, or other adverse events.

No one want to have a camera installed at home to be monitored 24/24. However, something that could check your pulse or maybe if you have moved in the last 30 minutes remotely would be much more acceptable.

It is probably a much bigger market than law enforcement.

In more institutional places, it could probably be used in hospital, nursing home, psychic wards or prisons.

I'm not sure why you need to be able to track bodies through walls to accomplish that. A simple ultrasonic sensor array in a room with some software could accomplish the same thing without the dystopian potential. I just can't see this not getting horribly abused.

> Gold Fig Labs: The startup is building a tool for version control on settings pages. The founders come from Firebase, where they were both early employees. The company has signed up 60 companies in the last five weeks, including “multi-billion-dollar tech companies.

I've been setting up some pipelines at work. I was thinking recently, "huh, these pipelines aren't easily version controlled at all". Unless you use something that is completely configurable by text, which is pretty hard to accomplish, then it's basically impossible to get a snapshot of your services. It's even harder to roll stuff back. Or branch it off and test out a new configuration. I'd wager that a lot of issues in production are configuration related, simply because it's harder to introspect and code review versus code changes. Hopefully that can be fixed.

Maybe I don't understand what you're trying to achieve, but doesn't Gitlab CI fulfill that role? Your versioning system is essentially git. Branching is done trivially, rolling stuff back is running deploy pipeline from older branch or tag.

I'm talking about versioning infrastructure. So let's say you have a setup with a load balancer, three compute VMs and two database instances. You modify the config to make it three compute VMs and three database instances, but suddenly something goes wrong! Can you roll back your infrastructure easily?

This is a standard "infrastructure as code" task for devops/SREs/whatever , for ex using Terraform

Assuming you mean CI/CD pipelines, with at least Azure DevOps, Travis and AppVeyor, pipelines are configured using YAML files, which are themselves checked into source control (e.g. git), so they are versioned just the same as any other files.

Azure DevOps supports both a graphical UI and YAML files. Pipelines created with the UI are versioned within Azure DevOps, rather than source control. I get the impression that the UI is considered deprecated, even if Microsoft haven't explicitly said so.

More data processing than CI/CD. But really all infrastructure config should be versioned. Can you version an AWS Lambda config? I'm genuinely asking.

Not sure about AWS specifically, but on Azure you have several methods available to define your infrastructure as code (IaC), including JSON-based ARM templates[0]. AWS and GCP must have similar IaC capabilities.

[0] https://docs.microsoft.com/en-us/azure/azure-resource-manage...

While AWS services support multiple routes of control, infrastructure-as-code via CloudFormation JSON/YAML which can be managed like source code is one of the preferred ways.

>which is pretty hard to accomplish

At my last two companies we used Terraform and Kubernetes with everything stored as text in git. Has worked pretty well.

Holy grail, sounds like a moon shot. It got AI, robots, if it wasn't vetted by YC I would think it is some sort of vaporware.

On the other hand, if they can pull off automatic research and experiments at scale, it seems to be a real game changer.

> As privacy-conscious consumers speak up against the proliferation of facial recognition tech (...) Traces is building computer vision tracking tech that relies on cues other than facial structure

It seems that Traces has slightly misunderstood the concerns of privacy-conscious consumers, so I would like to publicly make the following clarifications:

1. If your technology can track me as effectively as face recognition does, then it is not in any way "less invasive". You are still invading my privacy.


What I think would be cool would be a prediction market for each year's YC startups, limited to people with HN logins. Be interested to see how the cumulative wisdom of people here compares to reality.

Every time I read one of these my heart freezes up at the possibility someone has beat me to market. Then I remember why I spent 3 years in R&D.

No one else trying the same thing as your startup is a far bigger problem than being second. If no one else is doing it then you have to single-handedly prove there is a market, spend the money to educate people that they have a problem worth spending money on, and do all the work to figure out what the problem is and how to solve it in a way that represents value to your customers. That's really hard.

Leaving those problems to someone else and being second with much cheaper solution that can undercut the company that was first is often a pretty good strategy.

3 years in R&D sounds very normal for hardware projects. That scares me a lot to start something really complicated alone. I persuaded myself, that it’s just hobby and crawling slowly forwards.

3 years for software is enough to 3 times succeed of fail.

3 years in r&d is called a masters. 5 years a PhD. Not trying to be glibe but get your business house in order.

I already sold 3 companies doing the agile thing, I'm quite aware of the known problem/unknown solution approach. This time I'm more interested in the unknown problem/unknown solution space and, for this particular project, that meant inventing a new branch of mathematics to solve a particularly hairy problem.

What are you building?

Don’t reveal all, just a hint, like the industry.

I'm interested as well...

In short, a non-static, continuous, non-deterministic neural network that can span infinite state by estimating an NP-hard convex optimization problem.


Interesting to see that they got 10,000 paying customers in 2 years.

here are all S19 companies http://app.crunchdex.com/s19

Dashblock and Actiondesk look pretty slick!

if you could invest in one startup from demo day 1, which one would it be and why?

Nothing to fight global heating.

make the heat go away!

If you haven't read it yet please try to read the section 'In Comments' from the 'Guidelines' at the bottom of the frontpage.


Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact