Apparently, "a silicone diode always drops 0.6v" is an oversimplification. Wanted to use a couple in series with a resistor to decrease a DC signal by a constant, but the drop varies ~20% within the signal voltage range 😒
> Athlon XP is retrocomputing now
I can't argue with this, but I'm not ready for it either.
Ok, fine, Claude is actually useful, at least while the cost is 90% subsidized by the VC money.
To be fair Rust's async ergonomics are really bad rn, but it's amazing I get to use the same abstractions both for high-load gRPC servers with tonic/tokio and for real-time no_std embed with embassy.
Interesting, good catch.
Yeah, looks like printf is broken.
How do you print it out, with "%d"? If so, it means your value is stored as an integer and won't work with "%f", which expects a float.
It's a valid value, +inf: flop.evanau.dev/float-conver....
Interesting 🤔 Try disassembling the panic handler function, maybe the constant it points to is being corrupted/overwritten. `2.0f32` is basically a single bit: www.binaryconvert.com/result_float... .
Want to be as analogue as possible, but it's really hard to justify soldering 12 smoothing capacitors when I can just have a running average in software 😒
Completely agree with the 'infinitely more readable' part, but the comparison isn't entirely fair bc the underlying semantics and safety guarantees are different. Rust basically guarantees you have no data races, with Go, you have to settle for "haven't seen one yet".
6 lines, 6 data races and 6Gb rss 🤷♂️
Does this mean burning the forest and boiling the ocean to run LLMs is *morally right*, utilitarianly speaking?
8gb I guess? Still huge tho.
