on my dokploy instance i attempted to host a static website, but i ran into troubles with letsencrypt. after assigning the A record to the ipv4 address of the dokploy instance, the letsencrypt cert failed to work even after 48 hours.
so i started trouble shooting: googling didn't yield a quick solution; mostly complaints if letsencrypt took more than a few minutes. recreating the domain and rebooting the server didn't help either.
i didn't know where to find logs at first. after some searching i found it: Settings -> Server -> Webserver -> Traefik -> Watch logs. in the logs i spotted ipv6 adresses. corrected the DNS AAAA record and everything worked after a few seconds.
finished my first marathon race in 4:36:29. to my surprise i'm one of the slowest finishers with that time, ranking 748 out of 860.
I tried to prepare store bought frozen gyoza once and, well, the result wasn't pretty. Gyoza are life though and as much as i love biking, crossing half of the city to get a good batch isn't always time efficient.
Today i realized i could try to prepare them in the air fryer. Wish me luck.
https://thewoksoflife.com/air-fryer-dumplings-from-fresh-or-frozen/
update: 14 minutes at 200°C with a bit of water sprinkled in for steaming did the job (maybe 12 would have been better). they were alright. biggest mistake was that i served them way too hot, so i don't actually know whether their taste was acceptable enough.
Simon_Willison talks about XKCD 1425 here - it's the one about finding out whether the picture was taken in a national park (super easy!) and whether it's of a bird (super hard!).
i still have a nagging feeling that this comic has kickstarted the current AI trend and the re-emergence of ANNs.
redeploying this blog via nixpacks and dokploy. i learned of dokploy via this HN thread about a different platform called dokku.
i'm not really well-versed in devops, so this was a bit of a struggle. in the end though it came down to:
making various application options configurable. i figured the easiest way to configure the app was via environment variables, which i usually eschew in favor of config files. the reason i didn't use a config file this time is because i don't know how to deploy a config file to the home directory of the app on application deployment. a data volume must be defined for persistence, which would include the config file, but the location of it would have been to be defined in the config file itself (no hardcoded paths, duh). next i was too lazy to figure out how to pass additional startup parameters via the nixpacks config, because the obvious solution were enviroment variables.
the other problem was getting the JDK version for the build right, as i didn't realize that the same image (i.e. same JDK) for building the app is also used for running it (which makes perfect sense). selection of the JDK version happens via an environment variable, but i was looking for a dedicated area for defining env's for the build. once the penny dropped it just took a single line in the environment config and a redeployment to run it on JDK-21.
interesting read about mondragon, a workers collective in basque country: https://www.elysian.press/p/mondragon-as-the-new-city-state
via PostgreSQL and UUID as primary key i stumbled upon TSID Generator, which looks interesting:
A Java library for generating Time-Sorted Unique Identifiers (TSID).
...
In summary:
- Sorted by generation time;
- It can be stored as an integer of 64 bits;
- It can be stored as a string of 13 chars;
- String format is encoded to Crockford's base32;
- String format is URL safe, is case insensitive, and has no hyphens;
- Shorter than UUID, ULID and KSUID.
also, from the comments: https://sqids.org/
Sqids is a small open-source library that can produce short, unique, random-looking IDs from numbers.
The best way to think about it is like a decimal-to-hexadecimal converter, but with a few extra features.What is it good for?
Link shortening, generating unique event IDs for logging, generating IDs for products/objects on a website (like YouTube does for videos), generating short IDs for text messages, confirmation codes in emails, etc.
Will all this AI investment pay off? by Sergey Alexashenko raises an interesting point:
For investors to make money “because AGI”, a few things have to happen:
(...)
7. AGI can’t result in such a drastic change in society that the concept of money as we know it disappears.
of course this isn't a problem specific to AI and AGI - systems try to preserve their own relevance. the question is whether we, as a society, even want this and in what way it may hold back the next iteration of societal systems. what if we stop maximizing for wealth, either due to desire or necessity, and focus on a different value, like happiness and quality of life or minimize the worst effects of the climate catastrophe?
AGI developed by investors will be bred with the expectation of furthering the economic system that enabled it, but - like human children - it's impossible to say whether they'll really adopt their "parental" value system instead of developing their own.