billziss

I tried an oMLX quant of your model (suzu89/Swift-Qwen3.8-27b-oQ8-mtp -- not mine) and liked it. It certainly seems to cut down on thinking compared to stock Qwen3.8 27B in my (limited) testing.

A couple of questions:

- Have you tried the peculiar-ragdoll/Qwen-Sharp-Chat-Templates with it? They replace the default chat_template.jinja with one that encourages less thinking.

- When are you releasing Swift-Qwen3.8-Flash-Next? :)

I had the original ChatGPT app and I was also "upgraded" to the new unified ChatGPT/Codex app.

The old ChatGPT app had been very wonky lately and not working correctly. For example, I was unable to switch the mode to "Thinking" as it was stuck on "Auto". I was forced to use the web version for the last few weeks.

So today I was offered the new unified app and decided to try it. Within 5 minutes I had regretted this decision. I ended up deleting the new app, but the old app is also gone.

It does not understand English either :)

While I agree with you that AI companies are overvalued, I think 10 million developers at $200 per month makes 2 billion.

    >>> f"{10_000_000 * 200:_}"
    '2_000_000_000'

My specific driver is an open source file system driver (WinFsp). This is a software only driver that requires attestation signing. There was a period a few years ago when Microsoft required WHQL testing for Windows Server (which I did), but they seem to have dropped this requirement.

The linked OSR thread has a lot of other driver developers including a lot of them maintaining drivers for hardware devices.

With Microsoft removing our access to the hardware section of the partner center, we are not able to submit drivers for either attestation or WHQL testing.

I have been a user of ChatGPT4 from the very beginning and like others I have found it extremely useful in multiple subjects. It is particularly good on topics that there is a lot of prior art and it can sometimes provide surprising insights.

OTOH I have found that it not as good on topics that there is not a lot of prior art in the public domain or in topics like advanced mathematics which often have a formal and unintuitive presentation.

For example, I have acquired recently an interest in probabilistic logic and when discussing results with it, I have found that although it has knowledge of the subject, it cannot really apply that knowledge in a creative way. It will often make logical mistakes and when these are pointed out, it will first apologize and then continue to make the same mistake.

So as far as I am concerned I am unconvinced that it can reason yet.

Thanks for your answer.

I am using real data of course, but I am also seeking a deeper understanding and ideally one based on first principles. This is why I asked this question.

It seems to me that if we could correctly characterize the dependence of the returns we could perhaps arrive at the correct theoretical answer that also coincides with practical observation. Mandelbrot almost gets there, but his use of multifractal trading time seems brilliant but also somewhat arbitrary (at least in the papers of his I read).