I am surprised that vehicle manufacturers would support a weight-based fee, given that their most profitable vehicles like SUVs and pickup trucks tend to be the heaviest.

Intel’s doing interesting things with their Arc GPUs. They’re offering GPUs that aren’t super fast for gaming but are relatively low power and have a boatload of VRAM. The new B70 is half the retail price of a 5090 (probably more like 1/3rd or 1/4 of actual 5090 selling prices) but has the same amount of memory and half the TDP. So for the same price as a 5090 you could get several and use them together.

Is it feasible to run LLM inference comparably without CUDA or Rocm? How much of the cost performance goes away?

I have heard people claim the opposite: that Chrome is the memory hog and Firefox is much leaner. I think it’s probably dependent upon usage patterns, OS and extensions.

But I think the biggest problem with Firefox is Mozilla itself. I’d love to see a group with some actual backing behind them fork Firefox and make a proper competitor unaffected by Mozilla’s poor decision-making.

You could provide code to enterprise customers for a fee, with contractural restrictions on how they can use it.

You could also have trusted third parties see the code and vouch for it.

Or you may decide that the 5% asking for this feature aren’t worth it. You don’t have to capture every customer.

To me, the correct solution to the problem of being tied to one ecosystem crate for utility features like serialization or logging is reflection / comptime. The problem is not the orphan rule, it’s that Rust needs reflection a lot more than a dynamically-typed language does, and it should have been added a long time ago. (It’s in development now, but it will most likely be years before it ships in a stable version.)

If it’s just a Starbucks link every time I search for “coffee shop”, that’s fine. The one I hate is when Google Maps inserts ads into the directions it gives me.

“Continue straight for 0.5 miles”

“Continue straight, past the Pizza Hut, for 1 mile”

It would have been a lot easier to just say “continue straight for 1.5 miles”. It also makes me wonder if that really is the most optimal route or if it intentionally directed me past the Pizza Hut so it could advertise to me.

I believe they patented this as well, and it was pushed as a feature, "directions the human way."

The big thing for me was switching off LLM autocomplete. I found myself typing a few characters then hovering my finger over Tab while I waited for it to catch up, instead of just typing a line I already knew how to write. So now I only use an LLM by choice when I actually need it.

B2 has serious performance problems on the US West Coast (and possibly other regions too) in the evenings. Between the hours of around 5 pm to 2am, somewhere between 1% and 5% of requests transfer at <= 1% speed. This is bad because it’s very noticeable to my customers. It’s so predictable that I have a script running in a West Coast VM that pings me whenever the problem occurs - it pings me around 50 times a day.

Despite how predictable and reproducible it is, I’ve had a support ticket open for months with no progress. Having said that, even with this issue, B2 is still better than its (non-hyperscaler) competitors.

Yev from Backblaze here -> Do you have a ticket number I can reference and ask about?

Yes, because they would have to. Why start a pharma company for a disease that doesn’t exist? How else would you get people to invest in a company providing something people don’t want unless you have a plan to make them want it?

But then we should ban profit motivated medical initiatives, because even though it could bring huge benefits from time to time (ie a new medical procedure or a drug to cure existing diseases), it also bring raises the chances of medical disasters by providing a constant incentive for such entities to make disaster happen and profit from it...

In some states it’s easier now; in some it’s much harder.

https://www.brennancenter.org/our-work/analysis-opinion/chao...