But do you (or MSFT) trust it to do that correctly, consistently, and handle failure modes (what happens when the meaning of that button/screen changes)?
I agree, an assistant would be fantastic in my life, but LLMs aren't AGI. They can not reason about my intentions, don't ask clarifing questions (bring back ELIZA), and handle state in an interesting way (are there designs out there that automatically prune/compress context?).
What is her compensation ratio vs the minimum paid employee?
Are exempt employees compensated (stock?) in some capacity for being available at any time?
How do you monitor that managers are being good role models and/or helping to set boundaries? Without a company plan/objective, the only people holding the company to account is the individual, and they only power they have at the end of the day is to leave.
Is the Whitehouse fit for purpose in the modern age? Probably not. Is it a symbol of the country? Yes. Messing with that symbol on what seems to be a whim funded by corporate interests rather than doing something public and methodical is disgusting. Especially with a government shutdown.
We aren't even getting bread and circuses, just Nero at this point.
The actual White House, yes. Some out building of the compound, no. If you showed me a picture of it a month ago I would have no idea what it was. This whole thing is bribery, no doubt, but compared to all of the other Trump corruption this one is the least bad.
Are you asking how Dynamo compares at the storage level? Like in comparison to S3? As a key-value database it doesn’t even have a native aggregation capability. It’s a very poor choose for OLAP.
BigQuery is comparable to DuckDB. I’m curious how the various Redshift flavors (provisioned, serverless, spectrum) and Spark compare.
I don’t have a lot of experience with DuckDB but it seems like Spark is the most comparable.
BigQuery is built for the distributed case while DuckDB is single CPU and requires the workarounds described in the article to act like a distributed engine.
And yeah these days you can boost a single machine to enormous specifications. I guess the main difference will be the cost. A distributed engine can "lease" a little bit of time here and there, while a single RAM engine needs to keep all that capacity ready for when it is actually needed.
> treat self-hosting like a hobby and learn to enjoy it.
This is why I have stepped away from a lot of my self hosting. I have turned my attention/time elsewhere. Apparently though the time/money balance is shifting a bit again, so it may be worth it to go back.
My biggest hesitance to self hosting email specifically is dealing with spam. What does that look like these days and do you have any pointers to share?
> My biggest hesitance to self hosting email specifically is dealing with spam. What does that look like these days and do you have any pointers to share?
Postfix can easily be configured to reject incoming emails from senders without a reverse DNS mapping for their IP address, which makes it reject a lot of spam.
For spammers with reverse mapping greylisting still works fine, they almost never retry.
Certain commercial spammers (hello China :-0) use software which can be filtered with a just one rule matching their sending software, which is "nice" enough to display its name in their mail headers.
And last but not least spamassassin / rspamd work fine to filter whatever comes through.
In the end I get less than 10 spam emails per week. And these go into a separate mailbox filtered by good old procmail, based on spamassassin's ratings. I check the spam inbox maybe once a week for false positives and more often than not the box is empty.
> Postfix can easily be configured to reject incoming emails from senders without a reverse DNS mapping for their IP address, which makes it reject a lot of spam
Historically some corporate domains ignore that rule (yea, in 2025!), so I would advise not to reject any email and run everything through spam analysis daemon. This way you won't lose any email at expense of elevated load on your server
I use a combination of DNSBL and SpamAssassin. Nowadays Rspamd is supposed to be better than SpamAssassin, but SpamAssassin has served me well enough so far, and I haven't gotten around to trying out Rspamd. When a spam email gets past SpamAssassin, I copy it to a special folder, which gets processed by a cron job to train SpamAssassin on it (sa-learn).
Overall the mail server is very low maintenance. I had to add SPF and DMARC a couple years ago (DKIM isn't necessary) and integrate TLS with letsencrypt (just a few lines in a config file), and sometimes a Debian upgrade requires reviewing the configuration (several years apart as well). There's really not that much to do.
I’m not sure that there is any pre made product for this, but I’ve been playing around with LLMs to identify spam, or just generally sorting emails for you. And even the self hosted models seem to be pretty good at classifying emails even without external information like spam blacklists or IP reputation.
rspamd is my go to solution. Out of the box you get a lot of protection. I use Exim as my MTA but I suggest you use Postfix if you are starting from scratch, only because you will find a lot more write ups on it.
The biggest issue is getting an IP address which is not in the banned lists. IP reputation is key along with SPF and do not send spam!
In the UK a "business" static IP address is sometimes/usually/probably/might be OK. If you are unfortunate then it is already in the lists and you can check that out at point of sign up.
You might look into IPv6 too. I managed to do the Hurricane Electric IPv6 email thing on my home connection for a laugh. That was a few years ago. It seems I need to do something more to get to Guru status.
I've been lucky never to get very much spam to my self-hosted domain, but it went to zero once I implemented geo-IP blocking for a few obvious countries and has stayed that way ever since.
I don't develop for Apple platforms so I don't deal with Xcode (apparently thank god), but all of this reminds me of web development back in the early days of Internet Explorer. Until Mozilla/Opera/Safari/Chrome started taking market share did the development information/stability really start to drastically improve. I'm inferring that it's basically impossible to build tooling for Apple because it is all proprietary/undocumented, so their own tools have no competition. It blows my mind that you can't CI/CD builds on servers readily requiring people to setup weird mac mini server farms since the whole xserver line died years ago.
Ballmer's "developers developers developers" moment will forever be burned into my brain, but that wasn't wrong on a strategy level.
I agree, an assistant would be fantastic in my life, but LLMs aren't AGI. They can not reason about my intentions, don't ask clarifing questions (bring back ELIZA), and handle state in an interesting way (are there designs out there that automatically prune/compress context?).