>But empirically

Ok you said it. Where's your empirical evidence?

me: observes something

you: where's your evidence?

If you need evidence, look at https://www.levels.fyi/h1b/?q=anthropic . Anthropic seems to need to sponsor visas for their workforce as well. Just like every other large US company.

I was hoping for stats, not anecdotes. But thanks anyway

>The report finds that a broad path to citizenship would stabilize jobs and prices across the construction, hospitality, and care industries that families depend on while raising wages for all American workers.

The article posted is about suspending a Tech company's use of h1-b. The report makes no mention of the Tech sector and all the purported gains are in low-income high-physical labor industries like construction and hospitality. The whole wage and cost of living stabilization they're saying is due to immigrant labor is due to the fact you can exploit them for lower wages.

It further argues that Citizenship, not temp work visas, is the correct path. All while calling out that Tech companies, like Microsoft, are the primary beneficiaries of the current system.

In that light, I don't think linking this report arguing what you think it does.

You're at best naive if you don't see the big picture of sum of the immigration policies of this administration.

I would like to know the answer to this, too.

Not really sure what I should be doing on the side since AI seems to be automating a lot of work that I do as an analytics engineer. Maybe in 10 years all I'm doing is feeding in markdowns and project specs to an LLM and reading test outputs. At what point are you just a technician?

The bright side is that analytics engineer is a relatively modern job title, previously it was largely bundled in with data engineer, and before that, data scientist. So you're either already relatively early in your professional career or have adapted between programming disciplines once already.

I also (personally! I'm in data engineering for my part) feel it's one of the safer roles from an AI perspective. Yes, you'll be writing a lot less code in the future, but that's never been the particularly difficult part of your job: SQL is dead easy to use.

The difficult part is picking up all the business context and knowledge in order to work out the best ways of representing complicated and messy data for the business. AI can help to some degrees there, but if you're great at your job, your judgement is going to be built on so much business context that both can't be written down and would be too vast to feed into the model.

Sure, AI can do some of the job, but you'll always thrash it in terms of quality (whilst getting to spend less time doing the busywork of just writing your thousandth predicate of the day!). And worst comes to worst, all the business knowledge you have in your head makes moving into analytics a far more viable lateral move.

And that one is definitely AI proof. If all companies use AI for their analytics, they're going to be moving largely in lockstep with each other. Humans might be far worse, but often far better, so I'd be willing to bet that through sheer randomness that over time companies using human analytics will outperform those relying on AI.

We are mostly already doing that now as DE.

I just finished the Planet Money book. Pretty nice wide Econ 101 overview of how the economy works from a US point of view.

I'm going to read The Nerd Reich by Gil Duran next because that seems rather pertinent to the modern time. I haven't really looked at reviews and am going in blind.

$2000?! I know this is mostly sticker shock, but I wasn't expecting this price.

Looks like a cool device though.

Yeah, the COSMIC lack of polish caused me to jump to CachyOS.

I did something similar with an LED display on my computer case. The display required a proprietary UI program from the manufacturer to display GPU/CPU temps which unfortunately only worked on Windows.

Since it needs sensor data, WINE would not work here and I didn't want to do something funky with editing WINE or granting non-typical permissions.

I was able to reverse engineer the software using Claude, some Python, and a few hours of probing sensor data to understand how it worked and what was available.

I wrote most of the code myself (it was dead simple), but Claude was extremely useful in understanding what byte packets were being sent to the USB controller, what they meant, and what the controller was expecting.

I was able to make it into a service so now it Just Works(tm).

Probably the first time I've used it to "hack" something, but now I have a service that works great, I understand it, and I learned a ton about how Linux controls some low-level hardware.

Yeah, doesn't this reek of the Dotcom era where non-digital companies would just slap (.com) at the end of their name for the stock boost?

Is that why Anthropic recently gave out free credits for use in off-hours? Possibly an attempt to more evenly distribute their compute load throughout the day?

That was the carrot, but it was followed immediately by the stick (5 hour session limits were halved during peak hours)

> Is that why Anthropic recently gave out free credits for use in off-hours?

That was the carrot for the stick. The limits and the issues were never officially recognized or communicated. Neither have been the "off-hours credits". You would only know about them if you logged in to your dashboard. When is the last time you logged in there?

i suspect they get cheap off peak electricity and compute is cheaper at those times

That's not really how datacenter power works. It's usually a bulk buy with a 95th percentile usage.

I think it's a lot simpler than that. At peak, gpus are all running hot. During low volume, they aren't.

The advice to work on your own "shed" has really helped with my burnout during this AI age. I got into technology because I liked coding, building, and tinkering with systems. LLMs are great at coding and getting basic pipelines built, and I found myself with more pressures to be agent-first rather than hands on key-board at work. My side projects at home are where I've been able to find the joy for coding again.

I recently wrote a small service to get a temperature LED panel on my computer case working. It required a proprietary program to pipe the sensor temps to the display, which only worked on Windows. Being on Linux (arch btw :P), there wouldn't be a way to get this to work. I had a lot of fun learning about how to reverse engineer the inputs/outputs the other software was doing and replicate it in Python.