hietsu
@hietsu@sopuli.xyz
- Comment on AI Coding Is Massively Overhyped, Report Finds 1 day ago:
I disagree on the junk part: I see it so that if the output of the program are working, the logic must be flawless (just maybe not optimized when it comes to efficiency). Of course in our case the inputs are highly structured and it is easy for humans to spot errors in the output files so this ”iterate until outputs are perfect” has worked great, and yield huge savings in workhours. In our case none of the tools are exposed outside so in very worst case user may just crash the app.
But yeah I agree building any public frontend or anything business critical is likely the way to doom.
- Comment on AI Coding Is Massively Overhyped, Report Finds 4 days ago:
How is it not correct if the code successfully does the very thing that was prompted?
F.ex. in my company we don’t have any real programmers but have built handful of useful tools (approx. 400-1600 LOC, mainly Python) to do some data analysis, regex stuff to cleanup some output files, index some files and analyze/check their contents for certain mistakes, dashboards to display certain data, etc.
Of course the apps may not have been perfect after the very first prompt, or even compiled, but after iterating an error or two, and explaining an edge case or two, they’ve started to perform flawlessly, saving tons of work hours per week. So how is this not useful? If the code creates results that are correct, doesn’t that make the app itself technically ”correct” too, albeit likely not nearly as optimized as equivalent human code would be.
- Comment on Google's shocking developer decree struggles to justify the urgent threat to F-Droid 5 days ago:
We had a few good Linux phones back in the day but Nokia / Microsoft killed them trying to compete with iPhone OS and Android: Maemo / Meego were great but did not get a proper chance.
Jolla continued the legacy and Sailfish OS is still something worth checking out if you can find suitable hardware, or idk how complex it is to port it.
- Comment on Google's shocking developer decree struggles to justify the urgent threat to F-Droid 5 days ago:
Also, aren’t some critical apps like banking apps starting to ban unlocked / non-stock systems? Heard someone complaining about this a while ago.
- Comment on Microsoft Word documents will be saved to the cloud automatically on Windows going forward 5 weeks ago:
They’re prepared to do anything to get real user data for AI training. This little change gives them easily millions of files per day accidentally saved to cloud.
- Comment on YouTube just quietly blocked Adblock Plus — the internet hasn't noticed yet, but I've found a workaround 1 month ago:
Indeed. Tom’s Hardware for me has for long been one of the most useless tech news sites, mostly just dumb clickbaity ad articles in disguise.
If they would know anything about anything or done some actual research they would point to Firefox with a few relevant extensions that keep YouTube’s fuckery in check. Or the alternative mobile apps. Or stuff like Invidious. But guess they are too mainstream and thus afraid to upset Google in any way.
- Comment on Windows seemingly lost 400 million users in the past three years — official Microsoft statements show hints of a shrinking user base 2 months ago:
What is free though is LibreOffice, or some Nextcloud document addons (to a degree) if ”cloud” is the thing.
- Comment on Jellyfin over the internet 3 months ago:
Nice, but the bots may not understand the joke.
And not only that but they will tag the domain with ”there is something here”, and maybe some day someone will take a closer look and see if you are all up-to-date or would there maybe be a way in. So better to just drop everything and maybe also ban the IP if they happen to try poke some commonly scanned things (like /wp-admin, /git, port 22 etc.) GoAccess is a pretty nice tool to show you what they are after.
- Comment on Jellyfin over the internet 3 months ago:
Not at hand no, but I’m sure any of the LLMs can guide you through the setup if googling does not give anything good.
Nothing very special about all this, well maybe the subdir does require some extra spells to reverse proxy config.
- Comment on Jellyfin over the internet 3 months ago:
Use a reverse proxy (caddy or nginx proxy manager) with a subdomain, like myservice.mydomain.com (maybe even configure a subdir too, so …domain.com/guessthis/). Don’t put anything on the main domain / root dir / the IP address.
If you’re still unsure setup Knockd to whitelist only IP addresses that touch certain one or two random ports first.
So security through obscurity :) But good luck for the bots to figure all that out.
VPN is of course the actually secure option, I’d vote for Tailscale.
- Comment on Do you remember Windows 95? How about Windows 96? 3 months ago:
UTM is the way to go on modern Macs, and even iOS/iPadOS too! Built on QEMU and super easy to spin up virtual machines with any architecture.
- Comment on ELI5: How to put several servers on one external IP? 3 months ago:
Could be indeed. Looking at the nginx logs, setting a permaban on trying to access /git and a couple of others might catch 99% of bots too. And ssh port ban trigger (using knockd for example) is also pretty powerful yet safe.
- Comment on ELI5: How to put several servers on one external IP? 3 months ago:
I have wrestled with the same thing as you and I think nginx reverse proxy and subdomains are reasonably good solution:
- nothing answers from www.mydomain.com or mydomain.com or ip:port.
- I have subdomains like service.mydomain.com and letsencrypt gives them certs.
- some services even use a dir, so only service.mydomain.com/something will get you there but nothing else.
- keep the services updated and using good passwords & non-default usernames.
- Planned: instant IP ban to anything that touches port 80/443 without using proper subdomain (whitelisting letsencrypt ofc), same with ssh port and other commonly scanner ones. Using fail2ban reading nginx logs for example.
- Planned: geofencing some ip ranges, auto-updating from public botnet lists.
- Planned: wildcard TLS cert (*.mydomain.com) so that the subdomains are not listed anywhere maybe even Cloudflare tunnel with this.
Only fault I’ve discovered are some public ledgers of TLS certs, where the certs given by letsencrypt spill out those semi-secret subdomains to the world. I seem to get very little to no bots knocking my services though so maybe those are not being scraped that much.
- Comment on Jeff Geerling: Self-hosting your own media considered harmful (updated). Youtube removed his content, saying that self hosting content is "dangerous or harmful content" 3 months ago:
Saw the video… It mentions ”ripping” and even shows clips of some blockbuster movies. No wonder any copyright-sensitive automation gets triggered pretty fast. This will only get worse.