I would love to know how they are going to handle deepfakes.

Also, doesn't this give law enforcement an easier avenue to compel access to someone's accounts?

Yes it does. Biometric information (e.g a picture of your face) is famously not nearly as protected under the 4th amendment compared to things like passwords.

Anyone opting into this feature is effectively giving any US law enforcement (including ICE, local law enforcement, etc) free reign over their data.

Not just USA, same in many other countries too.

If a journalist or activist don’t use face or fingerprints to unlock devices.

> Also, doesn't this give law enforcement an easier avenue to compel access to someone's accounts?

It does.

They or the government is going to make deepfakes using this.

Would you accept 3% return on your money, especially if the investment was risky?

And your government can’t take it away on a Friday afternoon.

Interesting. I wonder if there is opportunity to train a set of small model variants to excel at a certain stacks. Eg Qwen3.6-27B for Node + React or Qwen3.6-27B for Rust + TUI

This is always how I've imagined small/consumer-hardware models going in time. If I only ever code in Python, give me a model that does just that (plus some general CS, algorithms, structure, etc.) and does it super-fast and well. Make it small enough that if I need a Python back end and an HTML front end, another specific model can load alongside and collaborate on the front end.

Or give me a pure shopping model that has a general understanding of products and product categories, and then will playwright/scrape/API into shopping sites to compare options and find me what I want. Etc.

prompting "no mistakes" was load-bearing

And if they do, their lawyers will use Claude to construct their legal case… which Claude silently nerfs as well.

"properly designed" - ah yes, programming languages are famous for universally agreed upon design philosophies.

Sure, but it did better on the test, which matches OpenAI's claim. More bang for more buck.

Interestingly, using your tests as a comparison, 5.5 low beats 5.4 medium at a 82% of the cost.[0]

[0]: https://aibenchy.com/compare/openai-gpt-5-4-medium/openai-gp...

5.5 is definitely smarter indeed.

And yes, 5.5 low seems to be as good as 5.4 medium, but a lot cheaper. In my experience that also holds up in real-life n8n agentic usage, so using 5.5 low is the best way to go atm.

I’m only seeing 3x the cost of DeepSeek for other providers on Open Router. Is there a better place to look?

I haven't really had this issue but deepinfra claims to have us servers and looks pretty cheap to me.