verdverm

the copy cats are getting tiring at this point

(my theory) they are not making money with it, so they save costs by using smaller models, resulting in poor results and feedback looping into fewer subscriptions

I'd recommend to start by using a platform for first experiments, like fine-tuning offerings, which can be done for lower cost and will handle the thornier issues therein.

Do some experiments to see if fine-tuning shows potential for improvements, curating the dataset and reward function are the harder parts anyway.

easy, none of them are good, they are all manipulated, speculation that looks more like gambling, and come with insider trading

chances are they will experience what StackOverflow did, though likely slower

I've pretty much given up on it as a useful platform, and with a login wall, I will not even be attempting to visit any more

they are pushing their Ai across all their web assets, also their own browser, with popups and notifications

very google-y to me

* Market, State, and Community: Theoretical Foundations of Market Socialism by David Miller

this is something that can be done with a cron, script, and LLM when conflicts happen - without the middleman needing to make a profit

this is how I have done it in the past anyhow (before project abandonment)

having your own hardware lets you do much more than serve LLMs, it's a great learning platform in general

the other thing is that the little models are getting better and more capable, so the hardware ROI becomes greater over time

try Jazz, infinitely more interesting than anything Ai could make

I saw a live band last week (daft punk meets prince), they wrote a song about Ai and used an auto tuner for it

> without relying on AI to do the writing for me

appreciate this if true, but don't put an Ai gen'd image at the very start then, it undermines the claim

It's been coming for a long time.

There are old HN submissions around home wifi routers being used for this. I recall one where a model was trained on video and signal interference, then tested on only signal interference. I bought a bunch of orange pi zero to do something like this, but then cryptos hit and my interests have evolved since (not crypto for long)

I won't touch brave or anything the crypto ecosystem

I'm using API search like EXA, the incentives feel more aligned

related: I created a generic testnet project that can support multiple interacting app views

https://github.com/verdverm/testnet

warning, unmaintained since I left atproto for various reasons

nope, having only exec is a terrible idea, that's where the teu any and everything leads to loss of control

I'm going the opposite direction to reign in their spaghetti throwing

we also have affection, which is really what they are coming for - with the chat bots that fake caring and emotions

one day we will call it the affection economy

It shall be interesting when the D party regains the house and can do investigations into all of this. We need to, deserve to, have some transparency. People feel the complete corruption, would be nice to have facts to go with it.

I turned off all Ai features, then DDG added their search summaries, despite my setting for no ai

DDG pulls stunts too

I'm moving to API based search options, using multiple, and rendering them as markdown for myself or my agent

on one hand, models are not as hard to train as they want you to believe, researcher churn between labs, know how diffusion

on the other, it's the same group, people connected to money, with soap opera drama

how long will that last and be given effort?

the direction they want to go is clear for me, similar to proton

It's really the worst of both combined. There is ample gambling like options and behavior on Kalshi

&&

policy makers, assistants, journalists, and the like are insider trading - and the president of the united states is intentionally market manipulating for profit

new lines between bullets, only italics available, not markdown

Founders stopped running it, same story as IBM

DDG is trying to be too much like Google now, with the Ai summaries and chat bot.

Actual search results are still bad. Seems to be a trend across most engines

For sure, definitions are abused and we've lost nuance. For example, it is possible to live in a democracy which has an autocratic president.

I started a book last night (Markets, , and Community) which critiques both extremes (libertarian, democratic socialism) on economic and political policy, then attempts to persuade the reader that Market Socialism is something we should try, living up to the professed aims of both.

> Live by market primacy, die by market primacy.

The market is not functioning well for the average person, so we petition the state to do something about it.

Live by laissez faire, die by laissez faire

the general issue is that many other people will be exposed through index funds that will have to buy into a dubious company

crop it out?

but that may only be the visible watermarking, what about the likely steganography like watermarking?

they are already doing this with text

best bet, stop using closed model Big Ai

Much of what DOGE did is likely illegal

scaling is hard, these issues will be revisited for those leaving github because of downtime