I just wanted to chime in and thank you for sharing your prompts like that!
It feels like which prompts people are using (even from developers on the same team) is often opaque. It's a great learning resource for people to see under the hood of each other's AI coding workflows, and I hope to see more folks doing this.
They definitely do train their own models, the founders have described this in several interviews.
I was surprised to learn this, but they made some interesting choices (like using sparse mixture-of-experts models for their tab completion model, to get high throughput/low latency).
Originally i think they used frontier models for their chat feature, but I believe theyve recently replaced that with something custom for their agent feature.
For come context, this uses a technique called CPPN’s (compositional pattern producing networks), where the inputs are (x, y, sqrt(x^2 + y^2)) and then there are several neural net layers with different activation functions and randomly initialized weights. It actually resembles the optical physics of light reflecting through glass (which explains why it sometimes looks like light reflecting through a gem).
At a startup I worked at, one of my cofounders was an incredibly skilled designer, and he took a lot of pride in the design of our homepage. He spent a lot of time in particular finding the right fonts, and would often complain when he saw other company homepages with inappropriate font choices.
One April fools' day, we thought it'd be funny to make him think that we'd changed the company homepage to have a comic sans font. Of course it'd be going too far to actually change the website publicly, so we decided to set up a DNS proxy inside the company office.
We took our company homepage, recompiled a static site with comic sans, and hosted it on an internal server. Then we set up a DNS proxy that resolved our company homepage for requests coming from inside the office to our comic sans static site, but otherwise the internet worked normally.
We made sure that everyone else in the company would feign outrage when he came in and checked the website, so the joke lasted for a couple hours before he thought to check the website from his mobile phone. Later, he appreciated the joke and figured he should have connected the dots sooner and realized that it was April Fools' day.
These are awesome! Thanks for putting your work out into the world, turtlesoup.
In my free time I make paintings of outputs from GANs/Deepdreams/etc. I've been wanting to paint something based on a model trained on illustrations, but I hadn't seen anyone decent GAN trained on that kind of dataset yet. Thanks for sharing your code, this is really exciting to see!
It's true that there aren't too many practical applications of GAN's yet, but I'd argue that transfer learning is already pretty powerful. It's fairly commonplace to approach a compute vision task by starting with VGG/AlexNet/etc and fine-tuning it on a relatively small dataset.
URX is looking for an experienced software engineer to take URX’s architecture to the next level by working on projects that touch all components in URX’s technology stack. As a polyglot organization, URX looks for versatile problem solvers with strong CS fundamentals who are excited to learn new languages and frameworks. The ideal candidate will not only build solid systems, but also excel at sharing knowledge with teammates and guiding the architecture of systems developed by the team.
=== Responsibilities ===
* Scale high-traffic web services
* Build distributed systems and data pipelines
* Design and own end-to-end, multi-tier architectures
=== Requirements ===
* BA/BS in Computer Science, or a related field
* Strong knowledge of Python, Java, Go, C, and/or Scala desirable
* 3+ years experience building distributed systems
* Experience designing, documenting, and building an API in a production environment
* Understanding of data flows and data integrity
* Interest in machine learning and information retrieval
* Experience with Lucene is a plus
ELI5 "what is a deep link?": mobile deep links mean registering custom URI schemes for your app, and registering a URL dispatcher within your app. The end result is that users can be sent directly to a specific page of content within your app. For example,
spotify:artist:0X2BH1fck6amBIoJhDVmmJ
...will take you to the in-app version of this experience:
The intuition is that sending users to in-app content is preferable to web content; IMHO, this is especially important for apps that have rich content or have logged-in functionality.
Example of a bad experience for apps that don't use deeplinks: have you ever clicked on a LinkedIn email on your mobile device and landed on a sign-in page in your mobile browser? If they used deeplinks, you're probably already signed into your LinkedIn app and wouldn't need to go digging for your password. This has been a common pain point in people we've talked to.
As a developer, allowing your app to be indexed will make it easier for other developers to send you traffic. Currently, we believe that the limiting factor preventing a "link graph" from existing across apps, is the challenge developers face in discovering how to find (a) which pieces of content are relevant to a page within your app, and (b) how to contruct a deep link URL to send users to those pieces of content.
Under Marissa Mayer's watch, Yahoo has been investing tremendous effort in refining the UX on their web/mobile properties like Yahoo Weather [1], Flickr [2], and Yahoo Sports [3].
Seems like they're giving people a reason to use their products, and then making a play to have an OAuth-level relationship with the users. Maybe they want visibility into how a user authorizes third-party apps (FB gets an awful lot of insight from FB connect!) or maybe they just want to solidify their user lock-in.
Either way, I can't imagine they justified the engineering work for massive redesign on seemingly not-profitable properties like Weather/Flickr without having a solid long-term plan for how to capitalize on that.
It feels like which prompts people are using (even from developers on the same team) is often opaque. It's a great learning resource for people to see under the hood of each other's AI coding workflows, and I hope to see more folks doing this.
(Link for anyone who wants to check them out): https://github.com/minimaxir/ballin/blob/main/PROMPTS.md