phkx

please use https links

[1] https://github.com/uqfoundation/pathos

[0] https://latacora.micro.blog/2019/07/16/the-pgp-problem.html

I‘m left with the impression that both our descriptions of these protocols and services are in need of some more substantial backing.

Telegram has stuck with me as a red flag. Mostly because Signal, which emerged around the same time, apparently had the better tech and was open. Not sure whether that changed.

See https://www.cs.cornell.edu/jeh/

The data you work with should be stored externally (volume, database, accessed via API, …). You don‘t keep persistent state of your workload in the container.

My main concern is with the content, because we're more experienced with the failure modes of humans when interpreting information. I like the other approach shown in the post, which is to cite an AI as you would with any other source. Best include the prompt then, which is also the best way when citing humans. Knowing which model generated the content on what prompt would at least enable some judgement on biases etc. which are present in the response.

Looking forward to the first AI-only interview magazine.

I‘m wondering what tooling is heavily dependent on the length of the hashes. Potentially if you want to keep the size of the transmitted data small (at work, we once considered git as a versioned database for an IoT use case…).

I personally would like to toy around with some graph database for knowledge management but haven’t found the time to get into it, yet.

Other than that, it’s dendron (and git).

Is there any common tooling to do that kind of financial modeling other than spreadsheets?

If the goal is defined in a too narrow scope, i.e. your ‚dumb‘ definition of efficiency, the flexibility may be missing. Still, that particular goal may be reached efficiently.

So it’s not an issue with the definition of efficiency, but rather with scoping the problem. As the article states, it may not always be possible to scope the problem in an easily measurable way, hence optimizing for proxy targets.

https://jakevdp.github.io/blog/2017/12/18/simulating-chutes-...

The Wikipedia article states that hyperlocal is used for information, which is relevant to the population of some given community. Does this always refer to a geographically defined community or has it been extended to other logical communities, as well?

I must say, that I don‘t fully get that example with the prod account. Sounds like it is a larger place where the ‚ticket‘ went to some other team to e.g. change some configuration of the prod stage. If we‘re talking about changes to the code, there should be code reviews from which you get feedback and can learn from it.

Overall, there seems to be a lack of well-defined processes and basic documentation and little awareness for onboarding tasks.

Even if no one had done some of the things you were tasked to do before - could you have asked the other devs how they would approach the task? Are there some dailies in which you can describe what you‘re doing and get feedback from the team?

Going forward, maybe try and identify one of the more experienced devs to build a more trusted relationship and try to get answers to your open ‚basic‘ questions from them. If you‘re afraid of annoying that person, rather batch a few questions and ask them in one session than asking small questions all the time, I‘d say. Depending on how well that goes, you might go further and tell them about your situation as you did here.

Since there doesn‘t seem to be documentation of e.g. the release process, you could write some and get a review for it, saying that you‘d like to prevent similar mistakes in the future.

Finally, if you don‘t find a way within the current team and you start every day in agony, find a new place and prioritize one which feels like it has a more welcoming culture.

On a side note: I didn’t find a Python library for time series generation (not analysis). Something where you can build some models (e.g. loan, income, expenses) which depend on a common parameter (time) and then evaluate all your models for different values of the common parameter. Right now, I generate pandas series/dataframes and combine them afterwards, which also took some massaging of pandas (which I also usually don‘t use a lot).

In my case, I simply settled for the maximum of all the different periods I encountered and file originals by the year after which I can trash them and use the digitized versions for actually working with them (e.g. my tax declaration is now much faster to do). For Germany, I found that 6 years grace period should be fine. 10 years, if you're self-employed.

Some additions: I found that black-and-white scans at 300 dpi work for almost all documents, resulting in a small file size and decent readability. Occasionally I switch to gray and 200 dpi and rarely to color. After looking up how long the originals of different types of documents need to be kept for legal reasons, I settled simply for the maximum time (6 years in my case) and file documents in a binder, sorted by the year in which I can discard them. Then, at the beginning of each year, I can get rid off one section of the documents which are older than 6 years. There is a second binder for active contracts (insurance etc). As soon as one of them ends, it goes into the first binder. I‘ve started organizing my Downloads folder in a similar way - sorting stuff by when I think I can delete it (either because it‘s not relevant any more or because I simply never touched it), typically a few months in the future. Both systems have helped to keep the clutter low and.

Their source code is here: https://github.com/zerforschung/schnelltesttest.de

So for me, the article is not so much about how the numbers in the current pandemic play out. It's about how society treats their 'weak', because each of us could become one of them.

—-

SLA: service level agreement, values of KPIs promised to customers

SLO: service level objective, internal target values for those KPIs, typically slightly more demanding than the SLA

SLI: service level indicator, measured values of the KPIs to check against SLOs/SLAs

I currently want to do that for a service built on several cloud services with their respective SLAs and my approach is to go through the combined probabilities to get an effective error rate. That‘s a bottom-up approach. I‘d combine that we a top-down derivation of what SLOs are required from the business side. If the first number doesn’t fulfill the business requirements with some buffer, we‘ll need to redesign. How do others do it?

Another thought that came to me - we already know of a technique to store phase information from coherent sources for practically indefinitely, which is holography. Can anyone tell whether that would be an option for the use case?

edit: I should have spent a little more thought on this comment, but I mostly wanted to get it out of my head and maybe have others pick up on it. The issue with holography is, that the image is created by interference of a reference beam and the reflections from an object. Here, we don't want to image an object, but the light source itself. Maybe we can learn something about the light source when we have multiple holograms created with reference objects...