shanebellone

If the problem seems vague, then the solution is not for you.

The concept of a "mom test" is condescending as well. Many moms code.

Yes, but multiplexing does minimize the problem. You could multiplex batched requests for hundreds of resources.

They don't reduce request count. For example, Chrome only allows 6 concurrent connections (requests).

If you load 10 images in 1 request, you've reduced network latency and avoided the connection limit.

As a publisher. One who seeks to increase load times on pages that require many resources.

I already have a working prototype that batches media requests. In my example case, I'm sending 1 request and receiving 6 images in return.

I'm familiar with css sprites too. That's still just 1 image per request thought.

Say you wanted to populate a social media feed... you'd need to dynamically call media. That produces a huge amount of requests.

What if you want to batch 10 images per request? Does anyone currently do that?

"I haven't understood why apps are not what you ask, so I tried to remember some well-known technologies/approaches."

I'm talking about websites that function like apps but have the privacy/security benefits of being sandboxed inside the browser (accessed by URL or link).

It seems we may be talking about different things. I am not advocating for downloadable apps (I've never liked the app store paradigm nor the access it grants random developers). I'm advocating for more capable websites (which IMO "ought" to protect user data--not leverage it).

Why such archaic examples?

Imagine visiting this website, Hacker News, to read a few posts/discussions. Does the idea of circumventing a refresh offend you? You could click into a threaded discussion (from the homepage) and back without triggering a reload or needing the data beforehand.

Other than the indexability problem, it seems like an ideal solution. In many use cases, the indexability problem is actually a feature too.

Ah, but that's not that case here at all. It's not an app. It's still a sandboxed web page. The only difference is the communication method. Rather than a transactional client-server relationship, we adopt the media-steaming approach post-load.

That is to say: You would load your desired website (the standard way) and then experience a "refresh-less" session for the duration of your visit.

It does sound like it, but no it's just plain JS. I recorded a short clip; showing is better than attempting to explain.

https://www.youtube.com/watch?v=6eSDuAe3pHA

By snoop-phone, do you mean stalkerware? I have not heard this phrase before. Thanks in advance for the clarification.

Let me clarify with some context. I've been tinkering with an infinite grid concept that consumes a streamed JSON feed (plus a sufficient data buffer to hide any delay from the user) to create and then display content with the help of a JS factory. All related media is then streamed to the browser and lazy loaded when needed. With this setup, you can traverse a database without refreshing the page by redrawing the window.

The only potential drawback I see is whether or not search crawlers could index content that's introduced via JS after a page load.

Edit: It also appears to protect from scraping... so I suspect it would conflict with indexability. That's a pretty big downside if true.

Really?

Delete my account.

"State is fundamental to how these models produce coherent output."

Incorrect.

Partially, I suppose.

The risk vs. reward component also needs to be managed in order to deter criminal behavior. This starts with regulation.

For the record, I believe regulation of AI/ML is ridiculous. This is nothing more than a power grab.

"I never said LLMs are stateful."

You said it multiple times.

On that note, goodbye.

"Prompts very obviously have influence on the output."

The LLM is also discrete.

"We arent meant to know exactly what it will output because that’s what it was programmed to do."

Incorrect, we can't predict its output because we cannot look inside. That's a limitation, not a feature.

'If you have some data and continuously update it with a function, we usually call that data state. That's what happens when you keep adding tokens to the output. The "story so far" is the state of an LLM-based AI.'

You're conflating UX and LLM.

I view it as long-form autocomplete.

"On a very fundamental level the LLM is a function from context to the next token but when you generate text there is a state as the context gets updated with what has been generated so far."

Its output is predicated upon its training data, not user defined prompts.

"Copyright infringement is not stealing, full stop. It is just that, an infringement on a commercial interest that the government grants."

This perspective is incredibly obtuse. You use the law (that you fundamentally reject) as a justification for a semantic debate that no one is having.

Whether you label it as infringement or theft, the impacts are identical. Regardless of the terminology, you are withholding a payment that is owed for something you've unlawfully taken.

Check your footing, your moral high ground is somewhere in the Mariana Trench; alongside debtors and fraudsters.

LLM is stateless.

Regulation is the only tool for minimizing crime. Other mechanisms, such as police, respond to crime after-the-fact.

I expected them to recognize and assess risk.

Humanity isn't stateless.