HN user

jacinabox

48 karma
Posts1
Comments122
View on HN

If you think about it, those who shape mainstream opinion on operating systems were always mostly bought off, and invested in the mainstream operating systems. That investment meant that a niche operating system like Plan9 or BeOS was never going to see wide support from the opinion shapers, even if those OSes had some interesting ideas.

I just did a captcha the other day that asked the user to select which items can fit inside the sample item (which was a handbag). You'd think that a multimodal deep learning model could figure out what objects fit inside other objects if it's going to cure cancer or whatever, but no I'm assuming that it needs to be taught explicitly.

The problem with natural keys is that nobody ever says, "My bad, I should have spelled my name the same way on this form as I did when I registered for the service, I promise not to do it again." Instead they say, "No matter how I spell it, you should have still known it was my name!"

that's nice, but the US digital asset stockpile is not the strategic bitcoin reserve. The sentence that controls acquisition in the strategic bitcoin reserve is the one before it:

'The Secretary of the Treasury and the Secretary of Commerce shall develop strategies for acquiring additional Government BTC provided that such strategies are budget neutral and do not impose incremental costs on United States taxpayers.'

I wonder what budget neutral means. Does it mean that they have to find spending cuts in other areas in order to acquire additional BTC?

One would think that the difficulty of making a company profitable when it's training larger and larger LLMs, combined with the diminishing returns and model collapse phenomenon, would make it so that companies don't wish to stop training larger models. I assume that they continue training larger models because whichever company stops training larger models would fall behind in the race to win new rounds of funding, but if that is the case what is the ultimate valuation that these companies are trying to achieve being valued in the biliions already?

Diminishing returns means that the user gets less marginal benefit with each larger model, and the model collapse phenomenon means that models trained on new training data might be less good than older models. Have straightforward mitigations been put in place such as filtering out from the training data forums where users like to share AI generated content?

Yes routers learn details about the destination of the packets they carry, but this is not a problem for most purposes. The author goes on to make some vague insinuation that tls implementations are insecure but if they are the solution is to fix the implementation. The Internet is designed so that all types of implementations of networking software can communicate on equal terms which means that it is already an open protocol. L take

The C language has rules around 'effective type' which determine what object type a block of memory can have, while the C++ language has rules around object model which does basically this AND requires that a constructor is called on an object before it is properly regarded as being of that object type. In my opinion the reason why the C++ standard cares about object lifetime is because C++ structs can have reference members which are required to be initialized in any instance of that struct type. In contrast it's compatible with what C has of an object model to just say to language users "If an object is in static or automatic storage the constructor is called automatically but if an object is in heap storage it's up to the user to call the constructor themselves."

amateur C++ coder

Back in 2021 I guessed that Shopify was overvalued because it didn't benefit from the same network effect that Amazon benefits from. (Also I preregistered this prediction so you know I'm telling the truth lol no I didn't.) Shopify has an app that app lets users buy from web stores that are Shopify built if they have the address. This difference is reflected in the size and staying power of their B2B customer base.

Decentraland uses a peer to peer-user connection model, which needs 'number of users' connections per client and will never scale. The intent is to not have to trust catalyst server with user data, but that's silly; you are already trusting catalyst server since you must trust its claims about identity of users.

Most bitcoin have never been sold. Bitcoin is a commodity that is intended to serve as a store of value, and other commodities like gold and silver do their work as a store of value without an appreciable fraction of them ever being sold. So using the term market cap is not a misleading measure in relation to bitcoin; it is simply used by analogy to how it would be used for a company; it's not literally implying that all bitcoin is the equivalent of shareholder value.

If I remembered correctly Rails originated the concept of an opinionated framework, and everybody else copied it to some extent. Yet it's the only platform that the author disses. Maybe this has something to do with the suspicious stuff that happened in the Rails community a while back.

Relief check's fat, don't stack any sats / Work at McDonald's but I wear different hats / They said, the risk is all here / We could be jacked up, levered up better for here / I can't count the dollars I have saved / One by one they all just fade away

I have no memory of what a 'contextual model' is! So I can't comment on how interesting the paper is.

The author doesn't define it either. Define your terms.