554 karma · joined August 17, 2019
(1) Sending email is nearly free, creating devices is not. Not sure if the economics match.
(2) False positive filtering matter because spammer need to manually reply to emails, creating more work. But with these devices this is not the case, it's all automated. There is no need to filter out smart people.
> sent you a new device you must switch to a new device
> this *kinda* breach
> there is a manual inside your new box you can read that
Seriously?
Just like ML research, we just skip the p-values.
The same goes for medium.com. There are good articles on Medium, but 95% of it is low-quality or spam, so I generally avoid these links or come in with the expectation that an article is likely spam.
[0] https://www.youtube.com/watch?v=2FzyKtct4es
[1] https://www.youtube.com/watch?v=wEVYS7Ofbgo
I don't know any better place though. I find myself using specific small subreddits more and more recently even though I was never a big reddit user in the past.
The point isn't to get rid of humans or trusted 3rd parties. It's to get reduce the reliance on a single centralized entity that may be corrupt, inefficient, or malicious.
The same principle applies here as physical eviction is fundamentally about risk. There is an "auction" on the eviction and potential evictors can decide how risky the eviction is. More risky evictions will require a larger payout.
1. An incentive to evict someone. Money. Someone needs to get paid to actually do the eviction because they are taking on some risk - physical in this case. It doesn't matter how they get paid. A smart contract can do that.
2. "Social/Legal consensus" that performing the physical eviction is the right thing to do. The person performing the eviction should not get punished for it. If the state and breach of contract is transparent on the blockchain this consensus could be achieved.
3. A way to confirm the eviction to pay the evictor.
Yes, the enforcer is not (necessarily) part of the contract. But if the right incentive system is in place and there is consensus that physical enforcement should happen - anyone approved could take that job/risk and be the enforcer.
With such a system you may not even need a physical enforcer since the risk of not getting the tenant out could be tradeable (like an insurance) and the owner would get his rent even if the tenant cannot be evicted immediately. In case of the tenant, there would be huge negative financial incentives to stay.
You could setup a similar incentive system as smart contracts. If your tenant does not pay, you could get him evicted much fatser and automatically, because another smart contract can pay the enforcer. And the overall fees are subsidized and tradeable.
Of course we're far away from doing that and both approaches have pros and cons, but it's arguable that the "human-managed" approach is clearly superior.
We're far away from both because, like you said, countries, governments are regulations are not going away anytime soon. But I do think it's inevitable that things will change and "global programmable finance" becomes the standard. I think it will follow something like Planck's principle [0] - it's not going to happen because people are going to be convinced by crypto, but because the next generation will adopt crypto as more of a default.
At least for me, that's the reason I'm involved in crypto. I can't think of a tangible benefit I get from transacting in crypto, but I would much rather live in a world where countries, governments, and banks are less powerful and more transparent than they are now.
I think in 100-200 years we may look back and ask ourselves how we could ever live with inefficient and corrupt governments, financial, and legal systems that are based on geographical locality instead of them being internet-based. We probably won't be using the current incarnation of crypto, but I doubt the current institutions will survive, and crypto is a step towards that.
To run a Nomad cluster that is feature comparable to a Kubernetes cluster you also need to maintain Consul and Vault. With that it's no longer so simple to setup/maintain. A minimal Kubernetes distribution (k3s) is significantly simpler - a single command and you're up and running locally or on EC2.
The other problem with Nomad are unknown unknowns. The community has established Kubernetes as somewhat of a standard. This means you can find many pre-built applications (Helm charts) and solutions to common problems. With Nomad, I need to build everything from scratch. That's fine for extremely simple things like single containers deployments, but anything with more complex requirements, e.g. something similar to a small Kafka cluster, takes significant effort.
I found k3s + spot instances to be a great solution. It works beautifully and is fault-tolerant across spot restarts.
20 years ago this was simply funny. Nobody took it seriously. It even got an "awards" page.
These days, nobody would ever want to be associated with such a site out of fear of press and social media jumping at them and ruining their career forever.
Not here to judge right or wrong, I think it's somewhere in the middle, but it's an interesting observation.
For example, intermediaries such as brokerages and banks taking money from your investments and selling your data (and order flow in the case of trading) is clearly inefficient and unnecessary, even if you don't care about it personally.
Personally I'm not a fan of the company, culture, and the cult-like following Roam has. I hope somebody else succeeds and takes over the market and I'll move there. But so far roam is the best life/knowledge management app I've used, and I've tried a lot of them.
But I think it's too early to call roam, the startup, a success. It has some mindshare, but the product has no defensibility. It's not that complex. A competent developer can probably build a clone of Roam in 1-2 months of full-time work. A lot of copycats have started already and I wouldn't be surprised if in a few years it's not Roam that has won but just another copycat that had a better UI. Kind of reminds of how Slack killed that other app before it... was it... hipchat?
Because of it's popularity someone created an org-mode package called org-roam [1] that does some of the things that roam does, but in emacs. These are not officially related though.
Now I've dropped it and I'm back to vscode and Roam as my life management system. Maybe the time I gave it was too short, but I didn't feel like org-mode or emacs made me more productive - the opposite. I spent so much time fiddling with configuration options, trying to understand elisp, finding the right emacs/org-mode packages, and so on. Yeah, these things felt productive, but in the end they were are a waste of time that got in the way of getting work done. Also, let's not talk about the time where emacs just crashed and lost several hours of work. Apparently the default config (I used Doom) didn't think it was important to handle that case.
Perhaps things are different once you've use Emacs for 5+ years and are intimately with its ecosystem and packages. So, if you decide to go that route, it's a long-term investment. Don't expect any productivity gains soon.
And honestly, even after using org-mode for 6 months for pretty much everything in my life, I didn't think it was that great. Yeah, it's pretty good, but I don't think it quite lives up to the religious hype it gets from some users. People are going to downvote me for saying this, but IMO org-mode is a pretty neat but outdated piece of technology with most of its following coming from people who grew up with it and have used it for 10+ years. The power comes from their familiarity with it, not from org-mode itself. It's not that great.
The last two times I bought a new one I thought "Not happy, but let's give Apple one more chance because they used to be good and everyone makes mistakes" - Regretted it every time. Keyboard issues (went to the Apple store 6+ times), Touchbar issues and it's annoying as a whole, overheating, GPU issues with external monitors, Bluetooth issues, and more. Next time will be the first time (really!) that I'm not buying a Macbook anymore.
Kubernetes already exposes all the metrics, so if you run Prometheus + Grafana you get that for free. You can also find pre-built Grafana dashbaords for that by searching a bit on Github (the official Grafana dashoard sharing page sucks).
I've been operating Kubernetes clusters for years now, and most of these are completely irrelevant to me, dead, or I've never heard of them. Don't equate bad lists with complexity.
The only reason we're not using a managed solution is because we have some specific edge networking and multi-cluster requirements that are hard to do with a managed solution.