my favorite thing to do with tmux is using a sessionizer script[0] (credit for the idea/original implementation goes to ThePrimeagen). allows for fzf-ing my projects and creating a separate session for each one. especially handy for bouncing around when working on features/fixes that span multiple repos

for (neo)vim users, flattening vim splits and tmux panes into the same level for switching with ctrl+hjkl is handy too

[0](https://github.com/tolly-xyz/dotfiles/blob/main/.local%2Fbin...)

this can't be how i find out...

This alone doesn’t mean much, but if the signs start to compound…

this goes a bit further than the typical "how do you make meth" jailbreak. notably;

>915 files extracted from the Claude.ai code execution sandbox in a single 20-minute mobile session via standard artifact download — including /etc/hosts with hardcoded Anthropic production IPs, JWT tokens from /proc/1/environ, and full gVisor fingerprint

why is it further than a typical jailbreak? you can just ask about this stuff generally, as long as you slowly escalate it. I have done it with each new flavour of code execution for models

i have a truly marvelous story about interesting marginalia which this comment box is too small to contain

>No regression. No noise. Just compounding.

it seems like half the articles I click on here are generated, and in half of those, nobody in the comments seems to notice (or care)

this is what claude is doing when he gets stuck in a "you're absolutely right!" loop

will private packages hosted on pyx be available for openai to use as training data?

Dude, no. That’s not how licenses work. That’s absurd. You really think that whatever little code is in pyx is providing any material training edge over the code that OpenAI already has access to? Get off it.

any hope that this could be applied to improving memory fab yields and ease some of the capacity constraints on consumer devices? asking for a friend

Silicon has 23 known isotopes, and now you why it will unlikely ever be economical to reach 0 defects in a business context.

Modern chip designs do include over-provisioned features, so designers can often selectively downgrade areas that are not viable.

Chenming Hu books about solar cell physics and semiconductors are quite accessible. =3

For now, this is useful only for the processes used to make the latest CPUs, like Intel Panther Lake and Clearwater Forest in the 18A process and various CPUs for smartphones or computers that have been launched recently or which will be launched around the end of the year and which are made with the TSMC 2 nm process.

Memories use more mature fabrication processes, for which it is likely that electron microscopy already worked well enough.

The article is about a better method for processing the output of an electron microscope, which enables a better image resolution than in the past and the 3D reconstruction of the surface of the device. This is needed for the 2 nm/18A processes and their successors, for which the existing tools were insufficient.

Less likely than just inducing more demand from the AI firms

i've been thinking about this a lot the past few days, particularly about how coding agents can _help_ with understanding

i've been using opencode/opus to help with debugging lately, and it (he?) will happily dive into the source code of a dependency, the dependency's dependencies, and the C code that its binding to, all the way down to reading the libusb driver code and explaining what is going on where

whether or not i could have figured that all out on my own is beside the point; i wouldn't have take the time on a tight deadline to dig in deep. i would have done some poking and experiments and shipped a hacky workaround

does anyone have a link to the full news segment? it's not clear to me from the clip in the link that he's talking about Palantir; he makes multiple references to "this technology" without ever clarifying what "this" is.

sounds to me like he's talking more about AI disruption at large