Honestly, I’ve gone back to dvd rentals! CinemaParadiso sends me as many films and boxsets as I can handle by post and it’s the same price as a standard Netflix subscription…

And it’s in blissfully high bitrate 4k! Anyone else notice that even the high quality options of streaming services are compressed to hell?!!

One thing I really admire about isomorphic and DeepMind in general is that they really try and encode human ingenuity into their models. The only reason that AlphaFold worked wasn't because they had better hardware or more money to train models - it's because they used every bit of information they had access to.

The same goes for AlphaGenome - it only takes a few hours to train the model, but it's freaking amazing what it can predict.

Frankly, I haven’t been impressed with AlphaGenome at all. What use cases have had good results for you?

IMO, out of all the products with the “Alpha_____” branding, AlphaGenome is the most underwhelming.

This is simply not true, advances in basecalling and using Duplex give you >Q25

I really don’t understand why this social credit concept has become so widespread. It does sort of exist but not in the big government breathing down your neck kind of way.

Here’s an example - you can rent e-bikes and power banks at low cost, they are everywhere. If you don’t return an e-bike or power bank after using it, not only will that company charge you, but you’ll likely get blacklisted from renting more, even at other companies.

This is a good thing, no???

Well it depends what legal protections are there to back it up. If you get threatened with blacklisting without recourse, that seems like a bad thing.

Feels like its the same poeple saying locked down OS stores are great because security and chrome engine in all browser is fantastic because then everything is standardized.

But worse because here we are talking deep human topics.

Such long period of peace made people forget about how the world work. And clearly studing history in school ks not working. I guess in the US they are getting a new shot right now to remember.

That sounds awful. What if you have a dispute and they're the ones in the wrong? You can no longer just take your business elsewhere.

understandable that people might have fears like this since this happens a lot in the west in my own experience, but in china you can always argue with them and the customer service is usually way better, or you can drop the nuke by reporting to the authorities.

Yeah for many cancers, CAR-T is the cure which is insane and cool. We just need to make it cheap enough that everyone can benefit…

Haha this is cool, it reminds me of the simple.wikipedia.com alternatives to English language pages. Is there an about page linked somewhere? I’m wondering if the images are AI generated or if someone handmade all of this

I wonder - everyone keeps talking about brain drain and how impactful it is, but is it quantifiable?

I think most researchers who loose their funding choose to do other things rather than migrate.

I guess the tax data can provide some insight but do we trust the current administration's reports?

Soon enough it will be

In 6 months we’re gonna see a HN thread: “I bought ask.com for £250k - here’s what I did with it”

250k would be a steal! More like $20M+

Make an offer to IAC they might sell it.

These kind of project reports showing consistent breakthroughs and clearly a finger on the pulse of what users are encountering as pain points are a good indication that the Asahi team are real pros :)

Look forward to switching back to Asahi full time soon!!

Honestly I've been doing a lot of image-related work recently and the biggest thing here for me is the 3x higher resolution images which can be submitted. This is huge for anyone working with graphs, scientific photographs, etc. The accuracy on a simple automated photograph processing pipeline I recently implemented with Opus 4.6 was about 40% which I was surprised at (simple OCR and recognition of basic features). It'll be interesting to see if 4.7 does much better.

I wonder if general purpose multimodal LLMs are beginning to eat the lunch of specific computer vision models - they are certainly easier to use.

I assume that by "higher resolution images" you mean images with a bigger size in pixels.

I expect that for the model it does not matter which is the actual resolution in pixels per inch or pixels per meter of the images, but the model has limits for the maximum width and the maximum height of images, as expressed in pixels.

Did you try the same with gemini 3 models? Those usually score higher on vision benchmarks