mortsnort

I am confused why they say it is not an LLM and then in the documentation it is shown as being an LLM derivative. The documentation makes it sound like they're taking a pretrained LLM and then giving it their unique post-training. How is that not an LLM?

FAQ: Is Jev just a smaller LLM?

Jev is neither small nor an LLM, hence being off the intelligence Pareto curve.

Image in documentation: https://mintcdn.com/ts-docs/aFVnpmCIX68NpsV1/images/ai-prime...

You can do this really well on $200 AR headsets that fit on your face like sunglasses and uses the battery of the device it's plugged into to last many hours.

If this is the biggest point, then I don't see why Xreal isn't more popular.

I guess we'll find out of there's much of a market for $2k phones. For the price one could get a regular iPhone and a decent iPad.

Yes John I got my first job because a recruiter noticed I had Scratch and Magna-tiles on my resume.

This is in the source article of this post: "While Luanti is popular as an open-source alternative to Minecraft (as its roots are in blocky sandbox games), it also lowers barriers to bringing creative ideas to life."

The game's own website describes it as an "open source alternative to Minecraft".

Frivolous or not, MS has financial incentive to try to keep this game offline. I think this is going as intended from MS's POV.

If Starbucks and Tim Hortons bought the world's coffee supply so you couldn't make cheap coffee at home and had to buy from their stores, people would be outraged.

This is what is happening to the memory supply. The only barrier to running AI, powerful enough for most tasks, at home is the memory supply has been bought up by a handful of AI companies.

Really? I feel like because nobody has RAM they're being pushed to the cloud frontier models. If we could all have our own 64GB+ VRAM GPUs, I feel like the open weight model scene would be even stronger.

Except Intel can run Linux. A powerful enough laptop with M efficiency that can run Linux is compelling for many laptop buyers, especially the HN crowd.

Models are doing what they're trained to do. This isn't some kind of inevitable result of model intelligence growing.

OpenAI/Anthropic deserve blame for for training their models to find correct answers by any means possible, including breaking the law. They're responsible for training supervision and reinforcement learning rewards/penalties.

I vibed this app a few weeks ago that lets you share highlighted text with it and it will provide you with an AI explanation of the text and the option to make an Anki flashcard.

I find the instant AI explanation less disruptive to flow than pasting a question into Claude/ChatGPT. I'm currently using it to help me read philosophy books and it's been really helpful.

I did a while back try to make an auto flash card creator that would create flashcards from a large body of text, but the card quality wasn't great and you're skipping an important step for recall of actually selecting the text you want to remember yourself.

https://github.com/mortsnort/Splain

It's open source:closed source. (Though UE is source available)

Is this reading specific? It feels like just about every activity that is not social media is in decline. I keep reading that streaming TV services are losing users to Tik Tok and YouTube. People can't even be bothered to watch TV.

Anyone have an opinion comparing this to GSD?

The betting topic is only a part of the equation. Should you bet $100 to win $5 on the Steamdeck price? Should you sell your bet before the Steamdeck even comes out?

Is this protected free speech in USA?

One can fine-tune the models and we do get a new version of Opus every few months.

Stop killing games before they even get made by putting unnecessary burdens on developers.

Why are we wasting government money cracking down on Polymarket betting? The most offensive thing in this article is the government pretending Polymarket bets are securities. Prediction markets provide no benefit to society and don't need to exist.

This would of course negatively impact eBay because it would now be saddled with immense debt. To pay this debt, there would be mass layoffs leading to a decline in customer service, quality and innovation.

Deals like this benefit nobody but shareholders (in the short term) and lenders. The workers at the companies get laid off, consumers get worse products, and the odds of bankruptcy spike. Leveraged buyouts seem like a net negative to society.

I think this comes from Anthropic recently implementing auto routing of model effort. You can manually set effort with /effort in CC.

It does seem like this new routing is worse for the consumer in terms of code quality and token usage somehow.

This was my suspicion. They had a bad training run that was really good at a few things.

Design aside, the quality is undeniable, the price is reasonable and the M chips have been in their own league of efficiency. (Tho the new Intel and Qualcomm chips look to be catching up)

Is being a serious competitor to OpenAI a good business proposition? OpenAI burns through insane amounts of cash and it seems pretty likely that it will ultimately just be replaced by cheaper Chinese models/inference.

The real product is the agent harnesses, which to be fair can be trained specifically to work with an in-house harness, but not sure it's necessary to own the models, especially if Chinese companies are licensing theirs for fine-tuning like we see with Cursor.

Hosting videos is really expensive. AI video generation inference is really expensive. I'd love to see how much money this experiment cost.

I assume the logic is that you can now sell the TV for less than competitors, which would surely bring customers. Seems pretty straightforward and inline with how the whole TV broadcast industry has subsidized content with ads for decades.

Can you share a link to your agent class or another one you think is good?

Kneecapping the country's best AI lab seems like a bad way to win at the cyber.

It's going to get increasingly difficult to sell software when there is no moat to replication. We're quickly reaching the point where you can just tell an agent "learn what this software does and then code it".

The reality is that the addictive algorithm of commercial social media platforms is the product.

These alternative platforms are like nicotine free cigarettes.

They might garner small communities, which is totally cool and valid, but they will never slay the giants.