This is interesting. Do you know what features the classifier is matching on? Like how much does stuff like whitespace matter here vs. deeper code structure? Put differently, if you were to parse the AI and non-AI code into AST and train a classifier based on that, would the results be the same?
Candidly, it's a bit of a black box still. We hope to do some ablation studies soon, but we tried to have a variety of formatting and commenting styles represented in both training and evaluation.
This is a really compelling idea – but I'm having a little trouble making the leap from the high level description to what it would mean for my projects in more concrete terms. Would it be possible to show off some example tests that the model generated and maybe even a story about how the generated tests caught a bug before the code made it to production?
Our landing page at checksum.ai has a video in the hero section of test. We added some graphics (e.g. the green checkmark), but the steps executed are real tests that we generated.
But the tl;dr is
1. We learn how to use your app based on real sessions (we remove sensitive information on the client side)
2. We train a model on this data
3. We connect this model to a browser and generate Playwright or Cypress tests
The end result is code written and Playwright or Cypress. You can edit and run the tests regularly
> To that end, I'd like a view where the focus is the art itself.
This is something we've been trying to figure out! If you have an account, you can view this: https://artinres.com/recommendations -- you're randomly shown one work at a time and you can like, dislike or skip. We use the data to recommend you additional artworks, and soon we're going to roll out a digest / news feed of new recommendations for users.
> I don't really know what I'm looking for in art. I just know some things I like, but probably not all. I love Monet and other impressionist painters, but I also love old Japanese woodblock prints like the great wave. I'm also a huge fan of cubism and I can't get enough of art nouveau. I'm not a huge fan of abstract art, but I like some. I love evocative photography, especially in black and white. I feel like maybe my taste profile fits a multi-spoke radar chart, where each data point is a relative preference rather than a binary I like this or that type tick box filter.
I'm very much with you on this. In my experience, taste has less to do with discrete, obviously-nameable qualities like a certain color or subject matter, and much more to do with the way the parts fit together to make a cohesive yet surprising whole. That said, we had to start somewhere -- and the filters have been pretty useful to a portion of our users.
Like you mentioned, as more people use the site, we're building up the ability to recommend art to people and we anticipate that being a rewarding way to discover new art.
> I love artists and design, and am shocked this has not happened yet, so Im rooting for you.
I really appreciate it! And I appreciate the thoughtfulness of your questions.
One aspect is that we're not business people looking at this problem in a cold, analytic light: we care about the wellbeing of artists and the joy that people get from having art in their homes. I know from my experience as an artist that most artists are suspicious of new businesses – but we've found a way to work with artists where they really trust us. We do more for our artists than just provide a self-service tool for them to list commodities and get some sales – we coach them on selling; we help them build community with other artists; we're building new features to fit their workflows; and we'll even hop on the phone with them and let them vent to us about life as an artist, haha.
Artsy seems like a great platform, but we're mostly targeting a different type of buyer than they seem to be. We're also working with artists directly, rather than through their galleries. So, despite both being online art marketplaces, I think it's a bit of an apples-to-oranges comparison between us and them.
Regarding the rest of the startup graveyard, when we talk to our users on the buyer side of our marketplace, they haven't heard of things like Saatchi, or even Artsy, in many cases. So I think that, to a significant degree, the mental category of "place to buy art online" is still open, at least for people who aren't already experienced art collectors.
So I think in our case the answer is timing + obsessive attention to our users.
My wife is a designer-turned-artist so I'm familiar with some of the comments above re: solo shows / the value of galleries. If you've read "The Case for the 12 Million Dollar Shark" then you must know that most people who buy art at high prices are wealthy, upper class folks who see art as much as a way to show off how cool / sophisticated they are and as an investment vehicle.
I've personally found the entire industry off putting at how much it's about prestige and how specific collectors see an artist or a piece.
But at the same time, I don't think normal people want to pay high prices for art. It just doesn't serve much value to them. If your users haven't heard of Saatchi or Artsy, what are they doing buying art and why do you think this is going to be a large market?
It seems like you'll end up with low priced artwork and artists who want to graduate to a "real gallery" when they get bigger, and so you really need volume to make to work. Respect the hustle from one YC founder to another but scratching my head on this one.
John, thanks for taking the time answer. I know artists are not suspicious but they will come around - thats not the hard part (tho it can feel harder than it should be).
Im not sure your argument that your buyers have not heard of Saatchi is a good thing (you're right artsy is different market). This means its going to take more education and cost for discovery. They should say I looked online and found these four places but dont like them because of X. Then you address X.
Not a noob question! It's a deep and interesting question. :)
Pricing art is harder than pricing most other things since artworks are, by definition, non-fungible / unique. This makes it hard for there to be a "correct" price. For the most part, we address this by letting our artists set their own prices (they have more information than we do.) Other factors that play in are the cost of materials and labor (often hundreds of dollars per painting) and the potential for the artwork to go up in price some day, like a stock.
> - I don't really care much about the Covid-19 relief. If you are donating then good for you and for helping the community. But I do actually avoid websites/services that try to advertise donations as a selling point. I can donate myself if I want to.
> - Also, do not penalize artists who do not want to donate to Covid-19. That's their business. They might be giving more, through other ways, to the community. They might be tight with money. Whatever. But the way the website is presenting it is that these guys are the generous guys.
More than anything else, we rolled out the 'COVID-19 support' features because of what our artists were telling us. We got flooded by artists telling us that they wanted to use their art sales to raise money for COVID-19 relief. But we also knew that many artists are tight on cash, so we said "instead of sacrificing your cut, let us give our cut, and you can keep yours during this insane time." The last thing we wanted was to penalize anyone.
Early on, when Art in Res was just an idea, we played with a model where we took zero commission and monetized other ways. But, when we talked to our artists about it, they were skeptical and would ask "but how will you make money?"
I hear you that, when looked at in a certain light, it might seems high – but we work closely with our artists and they seem to all like that our incentives are aligned. Plus, we do our best to use that 30% in ways that benefit them, e.g. by guaranteeing that they get paid if someone absconds with their work without paying it all the way off.
We take on the risk (it's one of the main reasons for our commission). If someone stops paying, either they send their artwork back to the artist or, in the worst case scenario, we pay the artist their full cut at a loss to us.
