You might be interested in Pixi: https://prefix.dev/
It uses uv under the hood for Python dependencies, while allowing you to also manage Conda dependencies in the same manifest (pixi.toml). The ergonomics are really nice and intuitive imo, and we're on our way to replace our Poetry and Conda usage with only Pixi for Python/C++ astrodynamics projects. The workspace-centric approach along with native lockfiles made most of our package management issues go away. I highly recommend it!
(Not affiliated anyhow, other than contributing with a simple PR for fun)
Messing with EEG headsets, like the ones by OpenBCI. Being able to perform some operations by just thinking about them sounds so incredibly cool - I've been thinking about this for several years, but it looks like quite an expensive hobby.
As an ex-EEG researcher I can say that acquiring an EEG device is probably the least difficult thing. EEG signals are messy, not reliable and hard to interpret. Without a solid foundation in cognitive psychophysiology you won't have a lot of fun with an EEG device. So, before investing a lot of money into an EEG device you should watch a beginner course on cognitive psychophysiology or get some literature to get the basics down.
On the other hand, you could probably hack a cheap ECG device to measure EEG signals.
Loved it! I immediately thought of a survey we do at work monthly as a check-in for HR purposes, to gauge how we feel about the company, our work, the workload and so on. Right now we do it with a simple Google Form with a mixture of a score based question and an open-ended one. I'd probably like it a lot more from an employee perspective if the questionnaire was served with a system like yours. I imagine it would be cool for a HR member to have a set of metrics derived from the conversation, like a dashboard. Just my two cents. The only negative point is that it feels like the conversation never ends - as others said, it felt like the assistant asked similar questions to points already established before in my mind. It'd be nice to have some sort of progress bar, not sure. Amazing idea anyway, I wish you luck!
I have been using this site: https://artificialanalysis.ai/ . It's still about benchmarks, and it doesn't do deep dives into specific use cases, but it's helpful to compare models for intelligence vs cost vs latency and other characteristics.
Looks nice! I was curious about the effort needed to replicate this result with the Google Wallet API and it ended up being a nice evening project to pull off. Sadly with a demo account you get an ugly [TEST ONLY] tag in the card title with a non published account, and the customization options of the card itself don't allow much leeway. One thing I miss is being able to change the font color.
Thanks carbonhell, yep you really are bounded by the customization. But if you really need one you can see a preview on the website and use code EARlY for the card to be free
The other day I found this: https://www.vantage.sh/
I haven't had the chance to try it, but it seems to cover all your feature needs and the pricing looks reasonable, perhaps check it out.
I've used Zola (https://github.com/getzola/zola) for a static project homepage a few years ago to showcase examples with a simple description and a wasm app embedded in the page, it worked perfectly for me and the docs was clear on how to use it. It was very easy to set up along with a GitHub action to automatically update the wasm binaries when needed. It is definitely a tool I keep in my mental toolbox as a good default.
Known password managers such as Bitwarden don't simply communicate the master password from client to server in plain text: https://bitwarden.com/help/security-faqs/, the master password is salted and hashed client-side, then salted and hashed again when stored in Bitwarden servers. Even if you managed to perform a MITM attack, you'd only be able to download your encrypted vault data, which would then require your master password to decrypt (locally). I believe talking about security consideration requires specifying a threat model, but for the average user such a setup would definitely be considered secure enough. A local only setup would definitely be more secure, but then as you said you'd lose QoL feature such as ubiquitous access, or nice UI/UX, no setup hassle, easy usage of hardware tokens and so on.
If one were to attack Bitwarden, he would either have to crack the encryption scheme to attack a specific user/business or target it through other means. Ultimately I think it's a small compromise of a small security sacrifice versus a big gain in terms of usability and availability.
It's my understanding, at least in the case of dashlane, that the master password never leaves the local device. It's not stored on the server anywhere, its not directly used as part of a log in. It is only used AFTER the encrypted blob has been downloaded to the device, to decrypt it.
