This is a really nice dev-friendly API, could see it becoming a compelling go-to for eg indie hackers who have moved simple view analytics to reasonably priced tools like Plausible/SimpleAnalytics over the years.

In some ways it feels like Loggly, worth taking a look at their query/dashboard system as it has some really nice tools for clicking around to slice and dice graphs using unstructured data.

Will have to give it a try some time!

Curious what applications for vector search you've found interesting

Does this have semantic search of some form? May be possible to implement all client-side with local browser models soon

it does not. i did look into it though [1] and at the time didn't find a good client side vector search lib. i wanted to avoid in-memory vector search since the size of the data can be significant depending on browsing habits. It is definitely possible though. I got a proof of concept working with victor [2] and client-side embeddings but it wasn't good enough IMO to ship.

[1]: https://github.com/iansinnott/full-text-tabs-forever/issues/... [2]: https://github.com/not-pizza/victor

You might consider making a tool for wireless internet service providers and mapping Crown Castle / other roof antenna space services. This would be really helpful in determining uplink reach (just may need some parameters on roof-top blocking structures)

ScreenStudio is really good, I use it for all of my capture. Main feature I find missing is ability to reorder or combine multiple recordings into one clip, or add audio from within the app.

You can do that with ScreenRun (which I developed) https://screenrun.app/

What bandwidth does a typical video call use per hour?

First TB is free. I would assume at least 1.5 Mbps for HD so ~5.4 Gbph ... ~.7 GBph so 3.5¢ per hour if my math checks out.

Unclear if that would multiply by number of callers.

Depends what codec! For 1280x720 with a VMAF score of 90 this is the numbers I use.

* H264 - 1.25Mbps

* VP8 - 1.00Mbps

* VP9 - 700kbps

* AV1 - 550kbps

Yes I love this approach, I use this exact format as a way to get ChatGPT to work with an entire multi file programming project in a single idempotent bootstrapping script. Then ask for changes to be given as the entire file again

better than chatgpt? Or canned because hallucinations were even worse?

It is really what you want out of a hacker-stuck-in-corporate story.

There was a tale maybe 10+ years ago about someone who automated their job with a script or Excel sheet or macro and didn't tell anyone about it. Having a hard time tracking it down again, anyone remember what that was?

Google reveals two things, followed by a bunch of shitty blogspam:

- Last year, there was a post on Reddit, with an HN discussion as well: https://news.ycombinator.com/item?id=29994776

- In 2017, there was a question about whether someone should tell their employer that they've automated their job: https://workplace.stackexchange.com/questions/93696/is-it-un...

I’ve found LLMs to be good for those sorts of vague “remember that thing like this?” searches. Bing (with GPT) or even Bard with potentially better search integration are both worth a try.

Lmao I love this. Is there a collection of these sorts of real-time experiments? Every so often a remote control robot to printer will come up on here/back in the day Slashdot/Reddit/Fark and it's loads of fun. I remember having a great time as a kid trying to flip the Plantraco remote control car by driving up the corner of the wall