varispeed

Since Epstein is no more, the governments became crazy about going after people's private data, perhaps hoping to find some spice. Like some politicians lost their source.

Shock and horror!

Does one drive cost a mortgage?

To make macOS remotely usable you need apps for switching windows, to reverse scroll wheel and many more. Ridiculous.

I am pretty sure they had solutions in their training data. LLMs only predict the next token.

> we called for pacing the frontier.

Translation: our models are getting shittier each iteration and we ran out of ideas. Let's invent scary stories and hope investors will lap it up.

Idiotic.

When the Astra one was last time run? It's probably better to run these 2-4 weeks after release when models get nerfed to get idea of performance closer to what it is.

More likely is that the solution was in the training data somewhere. Astra is incredibly dumb for a model - at least after it's got nerfed.

5.6 Sol is much more capable.

I don't think that's that.

Last time I went to a club and the place just stank of piss, poo and sweat. This was at so called world class clubbing venue in London.

Then the prices. Nobody can afford that.

Then people don't fancy drinking anymore and if you prefer weed, then likely it will get confiscated and you get kicked out.

Probably it is under Privacy because one day it will be transcribing everything you say and sending to government.

They already gave idea with client side scanning and now UK government will mandate that.

The next step is conversations.

They just want to cash in on high memory prices and prevent people from saving money.

Complete nonsense. RPi has not even given evidence of their claims.

They match my experience. Astra and Fable I rate below Sonnet. They are incredibly poor. They were excellent for a couple of days after release and then plummeted.

Maybe I am being routed to more quantised versions or less capable models with system prompt to fake Astra or Fable.

These benchmark are useless as they don't say whether they were done before or after Fable and Astra got nerfed.

Evidence that it is happening or is it RasberryPi gaslighting?

If someone sells RPi with 16GB and the RAM is working fine, then why would that be fraud?

If someone buys e.g. Dell PC, then adds 16GB of ram and sells it at 32GB without mentioning they put extra 16GB, is it a fraud?

That's ridiculous.

I stopped using Fable long time ago. It's worse than Sonnet. Opus is not much better.

This cycle of new model running at full quantisation and then nerfed few days / weeks after premiere should be called out. Anthropic should also drop the adaptive reasoning scam.

If I pay for Fable, I should get full, not nerfed model at honest pricing.

Regulators should investigate them.

OpenAI is no different. Astra has basically the same problem.

This is classic gaslighting technique corporations use to extract more profit. Modifications or upgrades made by small business = fraud, OEM selling different SKUs = legitimate business.

Don't fall for it.

> but seems to catch genuine fraud.

In what way this is fraudulent if device works as described?

No different than upgrading a PC and then selling it. You don't call it fraud.

> We therefore remove the commercial incentive by locking devices to their original RAM size.

Get fucked. Clearly you want to make all the profit and stop people from upgrading their devices.

I think this will go like this: - Resistance. People don't like being "secretly" filmed. - Most people will realise it does not affect their lives. - People will ignore it - New generation will accept as something normal to do. Everyone recording everyone.

Who cares if you can buy all the RAM available. To hell with small business and working class who now cannot afford it.

Stop hoarding RAM. Are they going to offload it through ebay? Unlikely.

At this point there is no point buying into Android if it becomes the same walled garden as iOS.

Likely this is how AI is going get us ended, not AGI.

With such an advancement of AI, can't parity be vibe coded?

You can try to use LibreOffice all you want, until you receive a form to fill where LibreOffice doesn't display checkboxes and recipient won't accept "x", because their automated software only uses MS Office and won't properly attribute it to questions.

You should never rely on just one model. Use Opus Max for review and Sol Pro for drafting, then Gemini, CoPilot for additional "junior" reviews. Once passes make no material difference, feed it again to Sol Pro for review. At the end still worth checking with actual lawyer, but this way you might get document as good as it can get. Depending on the domain of course.

(Fable and Astra are crap at these tasks)

I wouldn't trust any output from Astra unless they have not nerfed instances ringfenced for that task.

Tried again a simple coding task with Astra maxed out. I did it only partially, left memory leak and hanging pointer. Sol spotted are errors immediately. I put findings of Sol to Astra and it only done partial fix with the rest claiming will remain unresolved. Sol fixed it in a minute.

I then tried Astra on farily long legal document to produce an excerpt. The excerpt had wrong conclusion and logical errors. When pointed out that the source text has the claims other way round, Astra agreed and then when asked to produce the excerpt again, it again made the same errors.

Absolutely awful model.

This. Authy and a throwaway phone added in that sits in a safe place.

You never know that odd piece of metal that fell off the toy you bought 15 years ago will be useful in a few years.