> Hugging Face published a timeline that said the rogue agent broke into a sandbox — a testing environment — “hosted on a third-party provider’s infrastructure” and then launched its broader hack from there, Reuters said. Hugging Face didn’t name the provider, but Modal Chief Technology Officer Akshat Bubna confirmed one of his company’s customers was hacked, Reuters reported. Bubna said Modal’s platform wasn’t compromised in any way.

how bad does the app burn battery life for your watch?

I've been beta testing it and it seems like it's not bad for battery life at all, works like a charm

math notation doesn't bias towards English language understanding like pseudocode

but it biases towards the latin alphabet. so your point is diminished.

and Greek!

Token Type Price per 1M tokens Cached Input $0.27 Input $2.70 Output $13.50

Looks like they are first vendor to undercut in price.

Fired their pre-training team, shutting down Nova, shutting down the Adept Acqui-hire but still building another Foundation model

>According to the people familiar with the matter, resources have increasingly moved away from the existing Nova models and toward a new frontier-model effort led by researcher Pieter Abbeel, who came to Amazon through the acquisition of AI robotics startup Covariant. Known internally as Frontier Model Research, or FMR, the initiative has become a top priority this year.

Amazon (or at least AWS as I knew it) did this a lot. Multiple redundant efforts accumulate and silo under different leaders. Eventually a VP picks a winner and everyone else re-orgs or leaves.

Building a commercially viable foundational model business is one of the hardest things a company can do right now. Amazon's bread and butter was, is, and will continue to be building out the hyper-scalers. Amazon is actually well diversified already. They have a juicy stake in Anthropic over multiple rounds of investment since 2024.

Why is everyone's pricing the same? licensing agreements with moonshot as an inference provider?

> which OpenAI seems to allow via API but not Anthropic

Does openai still allow logprobs in their current gen models?

> And no, no AI company has ever come to us and asked to run training on all of our scanned copies

Yet

We actually tracked this over the last year, bedrock is significantly better than the anthropic direct endpoints

xAI did lose all of their founders and they are rebooting the company because it was "built wrong"