Cruise was actually just about to return to market after the October incident [1]. We had reached efficacy on all (much harder) internal safety benchmarks showing the car had significantly improved.

GM pulled the rug on us a day or two before announcing. The current Cruise CEO wasn't aware at all either. I have my own conspiracies of why GM did this, but GM also has a long history of fumbling the ball.

[1] https://www.cbsnews.com/sanfrancisco/news/nhtsa-robotaxi-cru...

[2] https://www.theautopian.com/here-are-five-times-gm-developed...!

It's just a hosted Postgres database, you could run this locally with Docker for example: https://hub.docker.com/_/postgres

You can do this with https://openrouter.ai/

but if you want to use google SDK (python-genai, js-genai) rather than openai SDK (If found google api more feature rich when using different modality like audio/images/video) you cannot use openrouter. Also not sure if you are developing app and needs higher rate limits - what's typical rate limit via openrouter?

also for some reason I tested simple prompt (few words, no system prompt) with attached 1 images and openrouter charged me like ~1700 tokens when on the other hand using directly via python-genai its like ~400 tokens. Also keep in mind they charge small markup fee when you top you their account.

Nope, this caters to shareholders

I've built (now improving) a fully automated real-time ML stock trader in Rust. It's hosted on-prem via 6 Raspberry Pis using Kubernetes.

Is it at all profitable? Worth the effort to maintain?

I've only done paper trading so far, but yes it was profitable.

I've learned a lot by building it, so for me the effort has been worth it.

I spy a paradoxical comment

Cleverness does not equate to wisdom.

How was this quantified? If there are numbers to demonstrate it's the most secure compared to previous years, why not show them? Qualitative statements with no foundation only draw speculation.

A more in-depth paper about this found the Swish activation often outperformed other functions: https://arxiv.org/abs/1710.05941

Most of the recent research is moving to GELU (Gaussian Error Linear Units) activation functions: https://arxiv.org/pdf/1606.08415.pdf

That's interesting. I didn't read the paper closely, but skipping to the pictures, it looks like ReLU, but smoothed out so the derivative is continuous. Intuitively, that seems useful.

I wasn’t aware of that one. Definitely interesting, thanks for sharing!