matteotom

I find it difficult to believe the inference only providers (Baseten, Fireworks, Digitalocean, etc) are all selling tokens at a loss.

Asking Claude for a rough estimate based on publicly available throughput and cost data for open weight models on modern GPUs suggests serverless, pay-as-you-go inference is profitable on owned GPUs with reasonable utilization (30-50%).

It seems the assumption you're making without justifying is LLMs will significantly reduce the cost of software development. Even if LLMs can reliably write new features (or even just fix bugs), the maintainer still needs to spend time (which is not free) verifying and code-reviewing the LLM-produced code.

I did. Your argument seems to be that LLMs allow users who want specific features to direct a donation specifically towards the (token) costs of developing that feature. But I don't see how that's any different from just offering to pay someone to implement the feature you want. In fact, this does happen, eg in the case of companies hiring Linux devs; but it hasn't worked as a general purpose OSS-funding mechanism.

Funding for open source projects has been a problem for about as long as open source projects have existed. I'm not sure I follow why you think specifying donations will go towards LLM tokens will suddenly open the floodgates.

I've personally run Gentoo with OpenRC+glibc and OpenRC+musl on my laptop. I assure you ditching systemd was easier than ditching glibc. The OpenRC system mostly just works (tbh thanks to a lot of great work by Gentoo devs). The musl system required probably a couple dozen patches to various packages to get a basic fully working desktop (most of which were relatively straightforward, but still needed manual intervention).

idk, the two finger "rolling" pixel-by-pixel scroll seems to work for me - Firefox (also foot terminal, Slack (xwayland), and Signal) on Scroll (a Sway fork) on Debian (testing) on a ~year old Thinkpad X11. I don't think I've done anything to configure or customize it either.

I got a Thinkpad (after a few years on a Macbook) largely because in the past the track point was a lot better than trackpads. But in those years, it seems hardware and/or software have improved enough that I barely use it.

At least for Python (since I'm more familiar with Python code and the Python ecosystem): progressive typing lets you incrementally add typing to an existing Python codebase. So you can have at least some of the benefits of typing for new or updated code without needing to re-write in a new language.

Were the mappings only a subset of the obfuscated classes/methods/etc? Basically making the mapping a sort of public API

My previous employer was about 1k engineers and elements of the interview process were research-backed, but it was mostly procedural (order of rounds, length of interviews, etc) and based on internal data. But I now work at a startup where we make a best effort based on everyone's previous experiences.

I ask this out of genuine curiosity and not to try to start an argument: could you help point us towards this research? This point frequently comes up when discussing interviews, but I've never seen references to specific research and I don't know where to even start looking.

What metric are you looking at when you say "the size of government has vastly and consistently outgrown the private sector" - AFAICT, excluding 2020 and 2021 (which I think is reasonable), the federal budget has been between 17% and 25% of GDP for the past 50 years (where the fluctuations are more a function of variable GDP).

The number of federal government employees has also remained mostly flat for the past 50 years (and IIRC most growth in overall public sector employment comes from schools).

I can't remember where I heard this but it stuck with me: Teslas are the new American muscle car - fast in a straight line, but otherwise poor build quality and lacking the attention to drivers' experience you get from European brands.

Kickback: usually refers to a company (or sales rep) giving a large "gift" to the individual at a company responsible for making a purchase decision (influencing that person to make a decision not in their company's best interest).

The Sopranos was a HBO show about a mob family. Historically, waste management in some cities has been controlled by the mob.

to be fair, I replied that I live in the US and asked for more info and it gave a lot more detail

Can you provide more details about said "significant evidence"? This seems to be a pretty popular belief, despite being contrary to generally accepted economics, and I've yet to see good evidence for it.