HellsMaddy

Suspiciously, Turso started having issues around the same time. Their CEO confirmed on Discord it's due to the Fly outage:

> Ok.I caught up with our oncall and This seems related to the Fly.io incident that is reported in our status page. Our login does call things in the Fly.io API

> we are already in touch with Fly and will see if we can speed this up

For some reason, the images weren't loading in the article, probably my adblocker. Here's the main image showing the process: https://web.archive.org/web/20241108024834/https://i.kinja-i...

Repo seems to be here: https://github.com/THUDM/WebRL

I wonder if it could be exploited via WebGL/WebGPU?

Honestly, even as a neovim user, I don’t find vi mode to be very ergonomic for interactive prompts, and I prefer emacs-style keybindings in these cases. The only time I feel the need for vi mode is when I want to copy something, but in that case I already have that capability through tmux copy-mode. I would prefer if the team prioritizes python-specific functionality first and foremost.

Why did you create this? Kids playing their music too loud for your liking? Couldn't you just... talk to them about it?

Tracy works fine on Linux! Just used it today.

I highly recommend The Autobiography of Benjamin Franklin. His writing style completely transported me to the time and place of early America. He led a truly interesting and inspiring life which I think HN readers would find fascinating. I think he embodied the entrepreneurial spirit.

Jq tip: Instead of `sort_by(.count) | reverse`, you can do `sort_by(-.count)`

This was my takeaway as well. None of the other retrieval methods this paper benchmarks against are specifically trained on the corpus. I think it would be more fair to compare "Self-Retrieval" against models which have been fine-tuned on the corpus.

This looks cool. If I have a log file with a format that is like `[timestamp] SEVERITY { json content }`, can I use the feature that pretty prints the JSON part, or does the whole line need to be valid JSON? If not could I somehow write a plugin that would allow me to parse the lines in order to accomplish this? That would be really useful.

Maybe one of those sites that come up when you search Google for 'track package' is selling tracking numbers?

I'd like to recommend https://github.com/ardov/huetone (https://huetone.ardov.me/)

It has a unique UI that I haven't seen before in similar tools, allowing you to see the relative L/C/H across all of the colors in the palette using graphs. It makes it much easier to create a palette that looks cohesive.

Neovim has an RPC API specifically designed to allow GUIs to render it outside the terminal. Neovide isn’t treating it as a generic TUI.

https://neovim.io/doc/user/ui.html

I built something like this. Luckily my monitor has an IR remote control, which allowed me to record and replay control codes to switch inputs. I used a USB hub that also had an IR remote control for my keyboard/mouse, and I used a teensy with an IR LED to send codes to both the USB switch and monitor. I used an IR repeater/extender to route the IR signal from underneath my desk where the teensy/usb switch are to the back of my monitor where it can reach the monitor’s IR receiver. I also made a little NodeJS app that shares my clipboard between my computers. Although the whole setup is pretty hacky, it works surprisingly well.

I was trying to find something about that, but wasn't able to. I would be very curious if someone else can find meeting minutes or mailing list posts where they discuss it.

I just did a deep dive into this (pun intended). The name appears to date back well over a decade (probably longer) and has roots in lower-level browser APIs like the implementation of `postMessage`.

> Structured cloning algorithm defines the semantics of copying a well-defined subset of ECMAScript objects between Code Realms. This algorithm is extensible by host enviroment to support cloning of host objects.

https://github.com/dslomov/ecmascript-structured-clone

Eventually, in 2015, it was suggested to expose the `structuredClone` algorithm as an API:

> Has anyone ever proposed exposing the structured clone algorithm directly as an API? Katelyn Gadd was musing about fast deep copy in JS on Twitter, and I proposed a hack to use postMessage to do so[1], which works but it's a little roundabout. Since structured clone is a primitive that the web platform is built on it seems like a primitive that ought to be exposed. I know this exists in other languages (Python has copy.deepcopy[2]) and there's an npm "deepcopy" module[3] with lots of downloads so this is clearly something people use.

https://lists.w3.org/Archives/Public/public-webapps/2015AprJ...

While this might sound like good evidence to support increasing your protein intake to build more muscle, it doesn't take into account the adverse effects of protein intake above the RDA of ~0.8g/kg of healthy body weight, unless you're over 65.

> The adverse effects associated with long-term high protein/high meat intake in humans were (a) disorders of bone and calcium homeostasis, (b) disorders of renal function, (c) increased cancer risk, (d) disorders of liver function, and (e) precipitated progression of coronary artery disease.

- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4045293/

High protein intake is associated with elevated levels of IGF-1, and "High IGF-1 levels increased the relationship between mortality and high protein":

> Respondents aged 50–65 reporting high protein intake had a 75% increase in overall mortality and a 4-fold increase in cancer death risk during the following 18 years. These associations were either abolished or attenuated if the proteins were plant derived. Conversely, high protein intake was associated with reduced cancer and overall mortality in respondents over 65, but a 5-fold increase in diabetes mortality across all ages.

> These results suggest that low protein intake during middle age followed by moderate to high protein consumption in old adults may optimize healthspan and longevity.

- https://www.cell.com/fulltext/S1550-4131(14)00062-X

Also important to note that one of the authors of the OP article is an employee of FrieslandCampina, the world's largest dairy co-operative and one of the top 5 dairy companies in the world.

I was thinking about how to mitigate this. First thought was to rewrite links for embedded content such as images to use a proxy server, like how `camo.githubusercontent.com` works, but this wouldn't prevent passing arbitrary data in the URL.

The only other things I can think of are to only allow embedding content from certain domains (the article mentions that Writer.com's CSP lists `*.cloudfront.net` which is not good), or to not allow the LLM to return embedded content at all (sanitize it out). This should even be extended to markdown links - it would be trivial to create a MITM link shortener that exfiltrates data via URL params and quickly redirects you to the actual destination.

They do: https://platform.openai.com/docs/guides/rate-limits/batching...

This reminds me of an idea I had for an OpenAI proxy that transparently handles batching of requests. The use case is that OpenAI has rate limits not only on tokens but also requests per minute. By batching multiple requests together you can avoid hitting the requests limit.

This isn’t really feasible to implement if your app runs on lambda or edge functions, you’d need a persistent server.

Here’s a diagram I drew of a simple approach that came to mind: https://gist.github.com/b0o/a73af0c1b63fccf3669fa4b00ac4be52

It would be awesome to see this functionality built into BricksLLM.

Works fine on Firefox on Linux for me.

Yeah, you will need to adapt the code a bit: https://gist.github.com/b0o/ac939eb1608f949765ac842fe28f53b9

I had the same gripe, and recently tried switching to dprint. It is much less aggressive with changing where lines break, and honestly it’s a huge relief to be able to break lines wherever I think makes the code most readable. It’s also significantly faster than prettier. I’ve been very happy with it.

Tested on Sway with both Firefox and Chromium. Resizing works (no surprise) but the position doesn't work. window.screenX/Y always return 0.

Nice! I love HTTP APIs like this with simple, user-friendly URLs like `https://${domain}/${query}`. Anyone know of a list of other services like this? wttr.in is one example I know of.

Norm from Tested has a short vid on Twitter of him in the sphere:

https://twitter.com/nchan/status/1722688222713749881

In Neovim, make sure your desired highlight groups have `gui=bold,italic`, e.g.

    hi Comment guifg=#7e7490 gui=bold,italic

In Kitty, set the desired fonts:

    font_family      IosevkaTerm Nerd Font Mono
    italic_font      IosevkaTerm Nerd Font Mono
    bold_font        IosevkaTerm Nerd Font Mono
    bold_italic_font Iosevka Term Slab

The terminal is by nature very hacky.

I've configured Neovim's syntax highlighting to make comments bold italic, and then I've configured my terminal (Kitty) to display a different font for bold italic text.

Using this approach, you can use up to 4 different fonts: One for normal text, one for italic, one for bold, and one for bold italic. And the font for each group doesn't necessarily need to be that style, e.g. you can use a non-bold version for the `bold` font, etc.