Farming GDP has grown 2-3x since the 1900s. It's just everything else has grown even more. That doesn't make farming somehow irrelevant work. There's just more stuff to do now. This seems pretty consistent with OPs point.

If you liked this, you'll probably also like https://ytmnd.com/

Biggest surprise here is that this website is still around. I was browsing it 20 years ago

Newgrounds still exists.

Having to pass a vote before you were uploaded and available for real I think was a good way to do things. Like a shitty peer review but for coolness.

Shameless plug of the YTMND site I made more than a decade ago: https://fvcks.ytmnd.com/

Such a simpler time.

Cool stuff, but why is there an entire loading screen and intro animation, then a play button, when it could just link to a gif?

Made me lose interest in browsing real quick

It's to make sure the GIF is in sync with the audio. It was a bigger issue when connections were slower.

I genuinely can't tell if this is satire or not.

2010 OG account. He's been holding in this opinion for over a decade, waiting for his moment.

Heh, my all time favorite: https://whatishoth.ytmnd.com

For me the process which works is the "make a mess, then clean it up" approach. Also known as https://en.wikipedia.org/wiki/Double_Diamond_(design_process...

Yeh, same. I also have a strong “do it now” attitude to try and reduce as much fluff as possible. I know a lot of people where they will add sending an email to the bottom of their very long todo list and it basically won’t get done. It takes two minutes, just DO IT when someone asks. In fact when someone asks me anything at work I drop everything, find the answer, and my next reply to them will be with the answer. I know context switching can be tough but other people are your portal to how your perceived in the outside world so they should be treated as priority. People appreciate this and you’ll be loved by your colleagues. Then you can go back to the optimisation rabbit hole you were half way down that probably wasn’t 100% asked for by the client anyway if you’re being honest.

My car can't login to my bank account.

Give it a few years. After all how will Tesla get that $99 every month for your self driving susbscription?

Your car and fellow road users' cars generally have your life, your passengers' lives, and other road users' lives in its hands while in use.

Well, I see how, especially for people who are close to death and want to provide for their loved ones, the answer to "Your money or your life" might lean in the other direction.

My car probably could be hacked to murder me in secret but frankly I'm not worth expending that kind of access on.

The threat model is really very different.

Why not just install it from Steam?

It is actually one of the most alive and welcoming spaces in London.

I will never contribute to a project that runs on this sort of ridiculous popularity contest system.

Hint: every software project at every company runs on this sort of ridiculous popularity contest system, the rules of the game are just not publicized.

Yeah but I get paid for that.

Why would secrets ever need to be available to the agent directly rather than hidden inside the tool calling framework?

Creator of Matchlock here. Mostly for performance and usability. For interacting with external APIs like GCP or GitHub that generally have huge surface area, it's much more token-efficient and easier to set up if you just give the agent gcloud and gh CLI tools and the secrets to use them (in our case fake ones), compared to wiring up a full-blown MCP server. Plus, agents tend to perform better with CLI tools since they've been heavily RL'd on them.

That doesn't add up to me at all. Agents are RLd on tool usage just as hard and you can provide an "authed API call" tool to whatever you want.

Token efficiency is a good argument actually.

Sometimes people are too lazy to write their own agent loop and decided to run off-the-shelf coding agent (e.g. Claude Code, or Pi in case of clawdbot) in environment.

Exactly.

> you should theoretically be able to take the output of the Lidar + Cameras model and use it as training data for a Camera only model.

Why should you be able to do that exactly? Human vision is frequently tricked by it's lack of depth data.

"Exactly" is impossible: there are multiple Lidar samples that would map to the same camera sample. But what training would do is build a model that could infer the most likely Lidar representation from a camera representation. There would still be cases where the most likely Lidar for a camera input isn't a useful/good representation of reality, e.g. a scene with very high dynamic range.

This is not true, there is just a specific protocol you have follow to do the building work. Yes it increases costs, but it doesn't explicitly prevent them.