tfsh

https://github.com/googleworkspace -> in the about links is also verified as part of the Alphabet enterprise - https://github.com/enterprises/alphabet

I browse logged out. Interact when them I do not.

The logged out experience is closer to the interests of the average person. So if you're not pruning (and savings) your interests, that's hardly surprising.

Because it's an acclaimed, often cited course by a preeminent AI Researcher (and founding member of OAI) rather than four undocumented python files.

I think I must have just spent more time (5 mins) looking at this repo trying to understand why you posted it, than you spent actually coding this.

I don't want to put you off, but there's no substance at all here, I'd have assumed Claude wrote it based on the fact you've vendored in rules, but the code is so questionable, even an LLM from 2022 would do better. E.g. 'flattenData' from utils could just be [1] rather than a BFS, though I don't really get why your public API allows TensorData to be a single integer in the first place, 50% of your logic is to work around that.

But rant over. My point is, maybe post this when you've built even 5% of PyTorch, or learnt something of value, or have something tangible to impart upon us, rather than a library of ill-thought-out array utils.

1: flattenData = (x: TensorData) => Array.isArray(x) ? x.flat(Infinity) : [x]

+1 to a Kobo, they cheaper and better than Kindles, with full Calibre support (https://github.com/kovidgoyal/calibre - OSS which has been in development for ~20 years!).

The way you install additional software is literally just moving files into folders whilst its plugged into your computer. I'm sure it could handle Tailscale.

https://news.ysimulator.run/item/121 - I was interested to see what the common archetypes would have to say about this very post, therefore I submitted it.

I've been PSAs before on the front page with a reminder to check your flagged stories. I and others visited the link and were surprised to see how many stories I had fat-finger flagged. In fact I had never intentionally flagged a story yet the list was at least 10-15 long

And why would cheaper GPUs damper the diminishing effect?

The author has a post about this; https://xn--gckvb8fzb.com/never-click-on-a-link-that-looks-l...

This is a punny code, and I'm fine - if not happy - that HN doesn't choose to render their underlying unicode symbols. It's very easy to spoof URLs this way, e.g. using a symbol from another language to craft a look-alike URL that can match a reputable site.

Browsers like now Chrome try and alert you if the URL visually looks spoofed (because they do support unicode symbols in the omnibox), but I'm yet to see how well this holds up in production.

And I hope in even 20 years we still can't use emojis here, our language isn't so pitiful that we must regress to brightly coloured symbols.

Any Googler can write code and open source it on the Google GitHub (within reason, the process is quite straightforward). So no, Google as an entity does not official endorse it, all it means is at least one employee is working on that particular effort.

Easier for humans to parse, but introduces the threat vector of malicious attackers modifying the history and force submitting malicious code at or before a pinned time. That's why lock files exist.

SHA is still the way to go for those who are security sensitive.

What a statement. Mozilla going bankrupt would be disastrous, I don't think you appreciate how much effort goes into maintaining and evolving browsers, there are very few entities well funded enough with the expertise to maintain a fork, and that's without making assertions on their altruism. Mozilla's is in the right place, even if far too often they miss the mark.

Google has systematically shittified the internet with Chrome by pushing bunk standards that other browsers are forced to adopt. Out of interest, what standards?

Right now there's no active public backends, there are a few prolific JJ contributors who are working on their own forge though (https://ersc.io).

At Google, JJ is natively supported for interfacing with the monorepo, this is where the proof of being backend agnostic comes from. Hopefully as the network effects catches on we'll see more and more of a desire for 3P forges.

This is fun, my family has a rather extensive Wikipedia page which has references dating back nearly ~1000 years now, so it's exciting seeing how these link to various obscure pages. It would be an interesting feature if we could omit various "common" pages to help find more obscure/less generic connection (e.g. broad supersets like countries).

Jane Street is a technology company first and foremost whose main product is high frequency training, which is similar to mang of these neo-hedge funds (e.g. HRT, Citadel). Which is contrasted to the more established asset managers (e.g. BlackRock, Vanguard, BRK, etc) which I can say from experience are corps tied down in bureaucracy.

This looks really cool, I'd like to give it a go. The idea of taking a screenshot of the terminal and then parsing that to determine the true colour support is definitely novel, though perhaps so, because for me I can't get it to work. Are there any debug flags I can enable?

So far it was able to take the screenshot correctly (https://ibin.co/8kaRr8TIanv2.png), however the parsing of that fails with the non-descript "Palette parsing failed." error.

Edit: enabled tracing at got this: https://paste.ee/p/ZyNxG9FK

It's not. You might be joking, but that comment still isn't helpful.

My understanding is this is part of Google's internal PSD offering (Public Status Board) which uses SCS (Static Content Service) behind GFE (Google Frontend) which is hosted on Borg, and deploys other large scale apps such as Search, Drive, YouTube, etc.

A JJ forge is the natural next step, and I'm excited, but posting such a link right now is premature. Are there any design docs, blog posts, etc, on how this'll work?

Perhaps you missed the associated documentation? This is a classification tool which requires input labels "uses an EfficientNet architecture and was trained using ImageNet to recognize 1,000 classes, such as trees, animals, food, vehicles".

The full list [1] doesn't seem to include a human. You can tweak the score threshold to reduce false positives.

1: https://storage.googleapis.com/mediapipe-tasks/image_classif...

I see this as addressing a symptom rather than the cause.

A direct result of technical staff leaving is the loss of siloed knowledge, instead of trying to address it at the final juncture, my opinion is that this type of knowledge should be proactively documented and shared with the rest of the team as the domain expert learns/works on it. That way, they're around to mentor other members, reduces the bus factor and when said person leaves, the rest of the team naturally continues their work rather than spending time on ramp up tasks.

I don't think you're wanting to converse in good faith, but on the off chance this is a question - yes, GCP was revenue losing for a number of years, but since Q1 2023 they've been profitable. It takes money to bootstrap anything - obviously - this is the case for the vast majority of companies and their offerings, especially so for one which requires vast amounts of compute resources, SREs, legal, etc.

To clarify this is managing my own library rather than buying books off Amazon. I tend to buy books locally and pirate the ebook copy which I sync from my computer. I don't feel inclined to line Amazon's pockets any more than I already have...

Massive +1 to this, I'd recommend using Calibre to manage your library and any Kobo reader as the companion, I manage a massive 2000+ book library like like this, and never encountered issues with syncing, corrupt file systems, etc, which I did regularly with Kindle.

Kobos doesn't require a subscription. I switched from a Kindle to a Kobo Clara Colour recently and it's honestly one of the best tech choices I've ever made. Kobos are hackable by default, so you can literally plug them into Chrome and flash new software onto them via WebUSB (or just via the file system). The real kicker for me though is the support with Calibre, I have a massive collection of maybe 2000 books, and this perfectly syncs with my Kobo supporting filtering, collections, etc. attempting this with even 10 books with my Kindle would routinely break down, books not appearing even after waiting hours for it to index, corrupt file systems, etc, the entire device is designed to push you towards Amazon store, including the scammy "pay £20 more to remove adverts" and them disguising the actual price as "reduced".

The fee GP pointed to is a monthly subscription similar to Amazon's Kindle book club offering.

But will this company be British or European? I'd love to think so, but somehow I doubt that. There just isn't the money in UK tech, the highest paid tech jobs (other than big tech) are elite hedgefunds but they get by with minimal headcount.

Exact same story here, spent 4-5k on my build with a 13700k which has blue screened hundreds of times in video games (R6, Hogwarts Legacy, Cyberpunk) over the last year (to the point that I don't even play competitive tournaments now).

I did all sorts, switching from Windows 11 to 10, buying new memory (twice!!), countless days debugging, updating my bios, etc.

I'm relieved to finally have found the cause, but my goodwill for intel is burnt.

Do you know what the general fix is, is it just a BIOS update?

I had about 1000 connections and found my linkedin homepage completely useless as it wasn't showing me anything about the people I care about (mostly uni friends, colleagues).

I went through and whittled the list down to about 150 people I would either consider a friend or.close colleague.

My homepage is still useless as linkedin shows me random trending posts, so that effort was a complete waste of time.

It hadn't occured to me that recruiters would see less than n connections as a negative trait. I'd expect the opposite, someone with 1k connections (e.g) seems like someone who connects regardless.

Of course it's an A/B test, not a manipulation or something nefarious. Just a team collecting data on whether the forums tab is useful/beneficial to the user.