A handful of super-rich families got together in the 90s, hired some people to put together a campaign to re-label the estate tax as the death tax and convince everyone it was causing families to lose their small farms, and we haven’t talked seriously about it since.
Larger scale family farms that would go over the estate tax minimums make up around 4% of all farms in the US, from what I can find. Disrupting about 4% of farms upon the death of the farmer does in fact seem like a bad idea to me. But thst didn't stop Stalin from liquidating the kulaks.
What if one person owned all the farms. It would be terrible if a larger scale family farm would go over the estate tax minimum, and would make up 100% of all the farms in the US. Disrupting 100% of farms upon the death of the "farmer" does in fact seem like a bad idea to me. Those kulaks must be protected.
I just have a desktop at home that I run inference off of. It is a great setup and I don't find myself wanting to inference models directly on my laptop.
That’s what I would do too, but I haven’t found a desktop build that can rival a Mac Mini or Mac Studio on performance per watt. I haven’t looked super hard, but it seems like Mac is in a different ballpark.
I mean, once it's a desktop, watts are pretty cheap, so it's a bit strange to optimize on that factor for the desktop form factor. For laptops it makes a ton of sense.
If I was optimizing solely for cost I wouldn't use a desktop in the first place. Watts are cheap, but they result in pollution and noise (e.g., fan) or you need some costly alternative. Also I just value efficiency.
Yeah, I've realized that the things I don't like in tech have everything to do with the culture and politics. When I've been able to work with a small team of people I really like and respect, I've generally been quite content.
I've rekindled by passion by working for a startup again.
My previous employer (which I also joined as a startup) ended up in a situation where the head product manager became VP of engineering (it's a complicated story - don't ask). We also had a yes-man director of Eng and together they went all-in on very orthodox scrum, where they sat in the sprint planning/point meeting and overrode every decision of what to take off the backlog and enforcing "themes" of each sprint to ensure that only product work got done. It was very rare that any tech-debt work got dealt with, and security work was only done if it burned down CVEs or other "quantifiable" metrics that were contractually obligated.
I ended up ok as there was eventually an exit, but the core experienced engineering team all left within 6 months.
Now I'm not only allowed, but encouraged to take initiative and while of course I do product work, I can also take a step back before taking two steps forward again.
Good for you. It’s so exhausting dealing with these people who are constantly chasing a fantasy that through some process, or through sheer force of will, we can achieve a system where all the feature work gets done super quickly and we never have to pause or slow down to handle engineering concerns because they simply don’t exist.
I think they get that control because they promise this nonsense to whoever hires them. Essentially, they're flim-flam men. They low-bid to get the contract and then try to make their underlings make good on the promises they made.
I'd never be hired as a CTO because I'd probably tell them "Yeah, you can have 95% less bugs, just stop trying to ship 100 features a quarter. Maybe do half that. Or you can ship twice as many features as you do now, just put up with way more bugs and instability! Instant "Strong No Hire."
do you mean "unorthodox"? What you describe sounds both terrible and not very scrum-like, at least ideally (I too have experienced when whatever terrible approach you use is labelled "agile" by leadership...)
I somewhat recently had a conversation about how we were going to start being more "strict" about how we do Agile (with a straight face). And they were right!
Well, except for the fact that they took over the planning process, everything else was orthodox. From the fibonacci pointing system to the retrospectives where we had to go into detail about how the timelines didn't line up perfectly. But we were "working faster" because we were gradually getting more points into a sprint! (queue eye rolls)
What's worse is that I kept getting written up because my main role was DevOps, which meant I was highly interrupt driven...which isn't something you can point reliably.
nothing tactical from me, but I've fostered a strategic approach over the years that's lead to a deep appreciation for the real-time experience. You can probably recognize when it's good (and bad) once you've worked for a while, and you really need to consciously pause and remark "If this isn't nice, what is?"^1 at those times when it is good.
A decade of consulting had me always ready to wrap my engagement at the end of any day, and (for better and worse) I carried this with me to future jobs. I always miss (at least some of) the people, but never the situation when it turns sour and I leave. The good news: you often get a chance to work with the good ones again (even if that's because you entice them away to your next gig).
Same here, it has always been a state transition. There's always that snake showing up who forces everyone to watch their back and eventually disband. A tech coop seems like a good option, but it's almost exclusively web dev.
Huh, as a Lanaboote user I’m surprised to see this on the front page. I use this in combination with sbctl for key generation. I’m mostly using it because I wanted to set up full disk encryption with TPM2 auth.
This won't help you right now, but GrapheneOS did recently announce a partnership with Motorola, so presumably in a year or so support will start showing up for some Motorola devices.
Side note: I did get the 10a on launch from Google Fi for ~300.
Here's the most relevant section I could find from the original source:
"Chrome extensions can expose internal files to web pages through the web_accessible_resources field in their manifest.json. When an extension is installed and has exposed a resource, a fetch() request to chrome-extension://{id}/{file} will succeed. When the extension is not installed, Chrome blocks the request and the promise rejects.
LinkedIn tests every extension in the list this way."
Hmm, can one fake-install extensions that randomly return yes/no to those queries ? It's pretty clear which files linkedin (and other sites doing the fingerprinting) is testing, one can observe it as the OP author points out.
It should also be interesting to see which other sites test those very same files, has anybody looked yet ?
"Chrome extensions can expose internal files to web pages through the web_accessible_resources field in their manifest.json. When an extension is installed and has exposed a resource, a fetch() request to chrome-extension://{id}/{file} will succeed. When the extension is not installed, Chrome blocks the request and the promise rejects.
LinkedIn tests every extension in the list this way."
Yeah, Gnome consolidates feature changes into major releases every six months, so it aligns well with fixed release distros like Ubuntu and Fedora that have release cycles that are aligned with Gnome's.
KDE drops a new point release with new features ~ every four months, and has a more flexible release schedule, so it is just to just get the changes when they are released.
I'm currently running KDE on NixOS unstable which is great, but if I weren't doing that I'd still be on OpenSUSE Tumbleweed.
Yeah, this is something I am thinking a lot about. Companies won't be able to sustain this level of spending forever, and one of two things will need to happen:
1. Models become commodities and immensely cheaper to operate for inference as a result of some future innovation. This would presumably be very bad for the handful of companies who have invested that $1T and want to recoup that, but great for those of us who love cheap inference.
2. #1 doesn't happen and the model providers start begin to feel empowered to pass the true cost of training + inference down to the model consumer. We start paying thousands of dollars per month for model usage and the price gate blocks out most people from reaping the benefits of bleeding-edge AI, instead being locked into cheaper models that are just there to extract cash by selling them things.
Personally I'm leaning toward #1. Future models near as good as the absolute best will get far cheaper to train, and new techniques and specialized inference chips will make them much cheaper to use. It isn't hard for me to imagine another Deepseek moment in the not-so-distant future. Perhaps Anthropic is thinking the same thing given the rumors that they are rumored to be pushing toward an IPO as early as this year.
reply