As mentioned elsewhere, the benchmark introduces bad pelicans in the training set. What I'm curious about however, if it's possible for a human artist to "poison" the benchmark by releasing some really good pelicans svgs and have all future models output their version.
They kind of already have a central point with 'ssh exe.dev', which hosts the interface for provisioning new VMs. But yeah, still one extra step for the user.
When I was very young I installed OpenSUSE on my underpowered windows PC, it was really a hacker man experience that is engraved in my mind as a core memory. As a child I just thought it was cool to have a new and faster desktop, but as I've grown older I've stayed with Linux for its ideas and principles. Hopefully these laws can be overturned...
I'm building a Typeform/Google Forms alternative that integrates into existing applications and stores data in your own database. It allows you to define forms in JSON Schema or JSON Forms. Forms can be added or removed dynamically, and submissions are sent to your backend and in turn stored in your database.
Built using our full-stack library toolkit Fragno [0].
I'm building a Typeform/Google Forms alternative that integrates into existing applications and stores data in your own database. It allows you to define forms in JSON Schema or JSON Forms. Forms can be added or removed dynamically, and submissions are sent to your backend and in turn stored in your database.
Built using our full-stack library toolkit Fragno [0].
I very much relate to the problems you're trying to solve, so much so that I have created a Stripe integration library [0] that has very similar DX to flowgrad! (not nearly as feature rich however). I've also written a blog post as to why I built it[1].
Key difference is where the final billing information is stored, our library also ships db schemas and the webhook handlers write to that local db.
You should also check out our (open source, MIT) library[1] this is built on, it could be useful for you guys!
I've compiled the "pelicans riding bicyles" benchmark into a single page[0], it only spans a year and not every model is exactly comparable but you can see clear differences between 1 year ago and today.
From what I've been told by former ASML employees, ASML is not particularly worried about startups inventing a better method than their current lithography process because they basically run the entire chip production pipeline. Improving only one step of this pipeline is not enough to break their moat.
I’m working on “Stripe Integration as a Library.” It seems that whenever someone uses Stripe, for example for subscriptions, they go through the same few steps: creating a database table, setting up webhooks, and implementing the events they care about. The challenge of course is that everyone uses a different stack.
I’m building this using our framework for stack-agnostic JS/TS libraries. On the database side, we currently support Drizzle and Kysely, with Prisma support coming soon.
https://bsky.app/profile/janschutte.com
https://janschutte.com/