myyke

There's this chart around gpt-5's hallu and error rates:

https://api.wandb.ai/files/byyoung3/images/projects/37269171/0da61431.png

from:

https://wandb.ai/byyoung3/ml-news/reports/GPT-5-Benchmark-Scores---VmlldzoxMzkwMTYyMg

I'm wondering what "response-level error rate" is exactly and it is measured?

gpt 4.1 says it's sampled production prompts, rated by humans. Is that it?

I use two Edge browser profiles simultaneously on Windows—one for personal and one for work-related activities, each logged into separate Google accounts.

I also have Superhuman and Vimcal configured with both accounts and open in each browser profile. In these web apps, links open in the correct corresponding profile.

However, when using the Windows native apps for Superhuman and Vimcal, links open in whichever Edge profile was last active, often mixing up personal and work contexts.

I can't think of a solution to this, but maybe I'm missing something? ― Is there a known solution to ensure links from native apps open in the correct browser profile, or is this a common limitation with native applications?

I'd like to start using a transcription tool like fireflies.ai regularly. But there are so many out there now (grain, notta, avoma, bluedot, tldv, ..., ...) that I'd love to get some guidance on which ones stand out.

What I'm most interested in:

- Stellar UX

- Very simple and reliable to integriate into all the video call platforms out there

- Great at deriving the real essentials from what was said: good summary and comprehensive action points (I don't care about the verbatim transcription (and think that should be commodity nowadays) so much and much more about the condensate)

- Good cost/value ratio (single user)

- Languages: de is a must in addition to en. Anything out there that can do swiss german?

- (Anything else I should find important but am missing here?)

Would love to get some input/opinions from people who are using one of these regularly.

Thank you!

I'm using a browser extension to download Docsend files that have download disabled.

It's working great but it's no longer available and I can't find an alternative.

⇒ Is anybody aware of either browser extension or Python library that can download docsend files?

Background: I'm scraping various URLs, but (as expected) encounter issues with some servers blocking the scrapes, leading to errors like timeouts or 403 forbidden responses. Currently, I'm using the requests library, but for problematic URLs, I've noticed switching user agents or using different tools like pycurl or wget can sometimes bypass these blocks.

Question: Is there a Python library that automates these workaround strategies, attempting multiple methods to successfully retrieve a URL?