On a side note, I find it interesting that a few recent projects are going for the Open Knowledge name. The Open Knowledge Foundation (https://okfn.org) is one of the first/largest proponents of the open data movement (think of it as a Free Software Foundation but for data, not software). They started in 2004 and developed many of the open data licenses and widely used infrastructure tools like CKAN (an open data portal platform).
Looks great! Just got a bit bummed about the pricing page. Our whole IT infrastructure costs $200/month, so "Enterprise-only" reads as "not for you" :/
The same company that supposedly views everything through the lens of liability and compliance does this:
> When the cars were operating under controlled laboratory conditions - which typically involve putting them on a stationary test rig - the device appears to have put the vehicle into a sort of safety mode in which the engine ran below normal power and performance. Once on the road, the engines switched out of this test mode.
>
> The result? The engines emitted nitrogen oxide pollutants up to 40 times above what is allowed in the US.
I wasn't involved at the time, so don't know. I just know that the cost was significant given the NGO budget, not only the initial development but also the ongoing support.
I feel as if we lifted an anchor that was slowing us down.
I am the CTO of a small NGO (10 people total, only 1 other junior Dev at the time). We supported two apps that were built by consultants. They were a mess. NextJS, React, about 4 micro services for a site that had 50 users per WEEK.
I configured a devcontainer with the old codebase and an empty repository and asked Claude to rewrite it as an old school server side rendered Django app.
Went to sleep. When I woke up it was 80% done. Spent another couple days prompting and reviewing and reached feature parity.
A bit later did the same with the other app.
Now both are deployed, reduced the server costs, complexity, and are orders of magnitude faster.
Without AI agents we wouldn't be able to do so (as usually is the case with tech debt).
AI agents are incredibly useful in this regard. Omarchy even releases some skills, so anything you want to configure is just a matter of asking the agent to do it.
Of course, it would've been better not to need it in the first place. However the experience is much better now (and Omarchy is great!)
> To be able to pay with Pix, one needs to get a CPF (Brazilian Tax ID).
It's not ideal, but you can use Wise to pay using Pix and India's UPI. You simply transfer the money from your local bank account to Wise and they transfer to whatever Pix you tell. It's almost instant.
Meanwhile, there are talks about integrating these systems. This is the obvious long-term game, a clear threat to Visa and MasterCard.
Right now in Brazil the only advantages of using a regular credit card are the cashbacks and convenience to use contactless. The convenience is going away -- Pix now supports contactless payment, but it's not widely accepted yet.
Yes. It can get a bit more complex as some otels require authentication. You can check Pydantic AI Gateway, Cloudflare AI Gateway or LiteLLM itself. They do similar things. One advantage of yours would be simplicity.
That looks great! Any plans on allowing exports to OpenTelemetry apps like Arize Phoenix? I am looking for ways to connect my Claude Code using Max plan (no API) to it and the best I found was https://arize.com/blog/claude-code-observability-and-tracing..., but it seems kinda overweight.
Congratulations on the launch! Is it possible to replay the tests against another URL? My use case is that I have a nodejs backend that I want to rewrite in python. I wonder if I could use your tool to record the API requests to the current server and use them to replay against my rewritten server to check if the responses are the same.
Another useful thing would be if I could create the tests from saved requests exported from my browser's network tab. In this case your tool would work regardless of the backend language.
I enjoy vcrpy and use it a lot, but it doesn't seem to be that similar.
Vcrpy is closer to an automock, where you create tests that hit external services, so vcrpy records them and replays for subsequent tests. You write the tests.
Here you don't write tests at all, just use the app. The tests are automatically created.
Thank you for sharing this. My grandpa passed away earlier this year at the young age of 97. We discovered a kidney cancer and decided not to treat him and bring him back home.
During his final days, he became unresponsive, only sleeping. The doctors gave us the option of feeding him through a tube. We made the hard decision of not doing it. Gave him all the medicine to help his body heal, but no invasive procedures.
We stayed by his side for the next 5 days. Playing songs that he enjoyed. Audiobooks that he loved. And just taking care of him.
Finally, his breath became slower and slower until it stopped and he passed away. I had the opportunity of being beside him during his last breath.
The passing of loved ones is always difficult, but I am grateful for how he went. He lived a full life and was incredibly healthy until the end.
Without knowing, we decided on a sallekhana-like process for him. It was the right thing to do.
I love this! I prefer digital stuff (less things to worry about), but I miss the physicality, especially when friends come over. Books or CDs become a conversation.
If you'd like to do something similar, but don't want to DIY it, check out Yoto Player [1]. This is a small music speaker and they sell a bunch of NFC cards to "play" them. You can also buy blank cards and use their app to add whatever you want to them (music, audiobooks, even audio recordings). It's really well made.
There are a bunch of other companies with similar products. Some use miniatures instead of NFC cards. If you search the web for NFC music player, there are a few FOSS apps on github so you can focus on the hardware part and use their software on a raspberry pi.
This is also great for elders.
P.S.: if you fancy a cool project, I'd love to see someone reverse engineering Yoto so it gets the audio from a local server instead. This way we can use their great hardware, but can use any NFC cards.
That's very interesting, although I'm still curious about the training resource usage -- not "only" inference. I wonder what is the relative importance of training (i.e., what percentage of the resource usage was in training vs. inference)
On a side note, I find it interesting that a few recent projects are going for the Open Knowledge name. The Open Knowledge Foundation (https://okfn.org) is one of the first/largest proponents of the open data movement (think of it as a Free Software Foundation but for data, not software). They started in 2004 and developed many of the open data licenses and widely used infrastructure tools like CKAN (an open data portal platform).
Nothing to add, just found it interesting.
Disclaimer: I worked there for a few years.