Liwink

open weight models are not free models. Inference providers should share a meaningful percentage of revenue with the model companies. In fact, open weight model seems a solid business model - they can leverage existing cloud infra for distribution without spending massive CAPEX.

It appears they just want to avoid responsibility for potential misuse in these areas.

But at the same time, IIRC, several major AI providers had publicly reported their AI assisting patients in diagnosing rare diseases.

I'd recommend one more step - after closing the laptop, bring a notebook and a pen with you.

People often get new ideas or unblocked somehow after stopping the work. If this happens, don't open the laptop again. Write it down.

What percentage of e-commerce will be taken by OpenAI by the end of 2026?

It appears to me that they are already well-positioned to become the next generation of Amazon with their current user base -

* AWS -> OpenAI APIs

* Amazon -> ChatGPT Shopping

I cannot image what if Tesla has a similar vulnerability issue, and someone took over all of its vehicles.. Or maybe someone is already able to do that, and just waiting.

Gemini 2.5 Flash is an impressive model for its price. However, I don't understand why Gemini 2.0 Flash is still popular.

From OpenRouter last week:

* xAI: Grok Code Fast 1: 1.15T

* Anthropic: Claude Sonnet 4: 586B

* Google: Gemini 2.5 Flash: 325B

* Sonoma Sky Alpha: 227B

* Google: Gemini 2.0 Flash: 187B

* DeepSeek: DeepSeek V3.1 (free): 180B

* xAI: Grok 4 Fast (free): 158B

* OpenAI: GPT-4.1 Mini: 157B

* DeepSeek: DeepSeek V3 0324: 142B

I'm usually the Anna in the group, and always appreciate being remembered, even though it's not easy for me to say no.

> But don't forget China still emit 16B compared to US who emit 6B tons of CO2

China has 1.4 billion people, while U.S. has 340 million people.