Our cat's are about the same age and one also started to behave differently. We kept a close eye on her and started noticing other issues as well like loss of balance. Turned out she had a bad ear-infection. Cats are really bad at showing what's actually wrong.
Could be she wants attention but doesn't know how to show what is wrong
Oh, ondoordachte praat hoor. Ik bedoelde het generiek als "één of andere fiscale constructie" zonder dat ik daar ervaring of kennis in heb.
Ik heb daar ook het vermogen niet voor denk ik, om de fiscale trukendoos open te kunnen trekken.
Gedoemd om te blijven en te werken voor de staat.
@Gubloinvestor Leverage is basically gambling. Similar experience here, did it once with a 3x leverage. Up about 60% in two days, flash crash during the weekend, all gone when market opened on Monday. Never touching that again.
Plain shares, buy often, sell only to reinvest.
That "user related information" can be a valid one though. If you persist logs for a long time, especially if you use a SaaS for log processing, and if it in any way ties user/usage info to an identity like a user's email.
Very useful for phishing.
It's also a basic requirement for GDPR if you store identifiable user data in log files.
Er zijn wel degelijk administraties die goed draaien en er zijn er die op hun tandvlees zitten wegens ondernemend of onvoldoende/foute middelen. Maar er zijn er ook, en veel, die hopeloos inefficiënt zijn of zelfs compleet overbodig, en die enkel nog bestaan onder Parkinson's law.
Is dat onkunde of onwil? Waarschijnlijk beide. Ik zie het ook in grote bedrijven: mensen normaliseren de inefficiëntie van hun werk omdat het de eenvoudigste voldoening biedt in een relatief inhoudsloze job: ze draaien hun uren in een collegiale sfeer - de prestatie is bijkomstig. Meer nog, veel mensen staan weigerachtig tgo verandering en automatisatie omdat ze bang zijn hun job of collega's te verliezen, en simpelweg omdat ze het zo gewoon zijn.
Een potentiële oplossing zou zijn om ambtenaren meer te roteren tussen verschillende teams en administraties. Een ambtenaar die 30 jaar op dezelfde dienst werkt is geen 'specialist ter zake', dat is een blok aan het been.
You could still be charged for reckless driving.
You see, even though there's no speed limit that doesn't mean you can just go stupendously fast. You can only go 'sensibly' fast, depending on weather, traffic, type of car etc. Which is obviously somewhat subjective, hence the investigation I reckon.
But anyway, imo the right decision. They obviously planned well and took precautions.
@ItsRobbAllen Now change the "private" into "public". You literally watch the people in front of you vote before it's your turn. Logically still exactly the same question.
Now change the question so you need to watch your kids vote first. Logically still exactly the same question...
@mattforney Now change the "private" into "public". You literally watch the people in front of you vote before it's your turn. Logically still exactly the same question.
Now change the question so you need to watch your kids vote first. Logically still exactly the same question...
@waitbutwhy Now run the vote again, but change "private" into "public". You literally watch the people in front of you vote before it's your turn.
Would that affect your choice? Logically it's still exactly the same question.
@waitbutwhy Everyone needs to vote. I have kids who can't read yet. My two year old's favorite color is blue. So I voted blue.
If you chose red you're an egocentric a-hole that thought this was an IQ test. Yes, pressing red makes it sure you always survive, but it also ends human society.
I used to have several repeating tasks as well that ran daily, but as my usage grew over time I noticed I got lazy in my prompting and was wasting a lot of tokens/context, so now I try to stay ahead of it by revising my workflow every now and then.
For repeated tasks I have Claude write skills and scripts to actually automate them almost entirely, and have Claude update those whenever a new issue pops up.
As for development I already used phased plan mode, which I extended with regular review/update analyses, but I also enforce plan mode now for every non-trivial task to prevent it going off the rails and burning up a window unexpectedly. Basically if Claude runs into an issue during implementation it halts and we go back to planning.
@PietSnot15 Los van het poeieren zijn de aardappelen uit de winkel ook gewassen. Haal ongewassen aardappelen bij de boer en je kan die maanden bewaren zonder dat ze schieten.
People posting these interview questions often fail to understand an interview goes both ways.
Though a question like this may provide some insight into whether a junior dev understands some of the basics, in most interview situations it would be totally irrelevant and even a potential red flag for the interviewee.
True, but also more a matter of perception I would think. Unit tests usually look useless because they are so trivial. You test if an invalid input always has the same desired outcome, while you know it can't ever get that input. But that's the whole point. Some time in the future someone could mis-use a method and actually provide it unwanted input.
The problem is that you would need to think of all possible wrong use-case up front which is impossible.
However the question now should actually be: do unit tests still make sense in the age of AI? At the rate AI coding is progressing it looks like package installs and reusing pre-compiled code will no langer be a thing. AI will just cherry-pick what it needs for the most optimal path.
Personally I do still use unit tests in combination with scenario-based integration test when porting or refactoring code, to have visual confirmation of the outcome.
If you do multiple similar projects you can obviously reuse a lot.
And sometimes you want to have it iterate over a plan multiple times even if all tests are green.
Make sure it builds a plan in manageable phases and keeps track of them. If there are more than three phases, let it double-check every third phase or so.
Don't start executing anything if the context is over 40% used, then you always do: clear context and execute ....
What works for me:
Step 1: ask Claude to setup skills, rules and permissions for your project. You do this per project because it totally depends on the type of project and whether it's new or existing. If existing let it analyse the solution; if new explain the techstack, overall project goal, conditions on architecture l, which types of tests should be included etc.
It will create a claude.md file with the project scope, and a folder structure under /.Claude with relevant skills, rules etc.
You then review these and adapt what's needed. It will also include a basic permissions file, which takes a bit of experience to get used to. You want to gradually extend this (auto+manual).
Step 2: go in to plan mode, but start by telling Claude it needs to persist plans + progress updates underneath /.Claude/plans.
Planning is key. You can have Claude add additional permissions up front which it thinks it will need, like fetching from GitHub or other resources.
Once you get the hang of it you can have sub agents assist in planning.
You want to know where the plan is - it always stores it but it isn't always easy to find where. Better to structure it like documentation in the repo.
(Oh yeah, I also have Claude write up documentation on everything it implements as part of the workflow).
Step 3: execute.
If you do get interrupted for trivial questions: answer but have Claude also update its permission file for the project or add a skill for it. Ask to explain first if it's unclear what/why Claude is asking. You can f.i. use /btw for that.