The IT Blog

  • GPT-Live … And the Engineering Behind It

    Finally I took the time to read OpenAI’s post about their GPT-Live Realtime system.

    If you’re interested in how to build such low-latency systems, I definitely recommend to check it out. But what I also had in mind while reading the article was the “no one needs programmers anymore in the age of AI” … Oh really? Well I don’t know how many people were involved in the whole project – and I’m pretty sure they used AI in the coding parts. But the whole process to get there, knowing how / where to measure latencies, how to circumvent them, enhancing the WebRTC protocol and designing a new, faster standard, checking geo-location impact, shadow testing, etc … Oh boy, do I see people talking and engineering tasks there.

    (more…)
    Fediverse reactions

  • How to configure LogRetention with journald

    Whenever you have a regular process, you definitely want to have some proper logging. But the very next thought after mentioning “logging” should be “LogRetention”. Especially when the process is executed automatically. No matter how often — as soon as the execution isn’t a one off, log rotation & log retention are a topic.

    LogRotate

    Especially on linux, it’s just so easy: use logrotate for example. Just one simple config file /settings and you’re set up with log rotation, compression and retention.

    journald & journalctl

    On my miniPC, I tried to avoid writing a lot of own log files but whent with journald and piped all logs into it. And never thought about retention. – Until today.

    I just set up another process and during testing, I noticed that I had more than 2.5gb of log files (journalctl --disk-usage) … dating back until very long. A quick check of the /etc/systemd/journald.conf confirmed that I missed some settings.

    journald has quite an elaborate set of options, just check the man page. Without big evaluation, I’ guess I’m fine with:

    Compress=yes
SystemMaxUse=100M
SystemMaxFileSize=5M
MaxRetentionSec=1month

    Also, don’t forget to reload …

    sudo systemctl restart systemd-journald

    And don’t spend a second thinking about overflowing logs any more…

    Fediverse reactions

  • My code vanishes from the web

    As a lot of people, I had most of my code published on GitHub. During my (reduce the amount of US services), I decided that I want to move my code somewhere else. Codeberg came totally handy and was recommended a lot.

    (more…)
    Fediverse reactions

  • (why) I do not like Excel / Spreadsheets

    It seems like I’ve earned a reputation at work for hating Excel.

    We recently had a workshop with different stations and flipboards about how to improve some processes etc. And well – I added a lot of “less spreadsheets” sticky notes. After a while someone approached me and asked what my problem with Excel was – if I maybe just don’t know how to use it.

    But no, my reluctance to spreadsheets comes from the very bad experience together with automation. Let me explain: Very often, Excel (or any spreadsheet) is used as a not-very-well structured database in some process. And whenever you try to add more automation to this process or want to re-use the data, we come to one of two scenarios:

    (more…)
    Fediverse reactions

  • Against Information Hoarding

    I am an information hoarder. I am somehow unhappy if I don’t get new information, new insights, new aspects. That’s not limited to technology but to literally everything.

    And since I do more and more self-hosting, I have two services on my candidates list: Readeck and Linkwarden. Then, with the rise of AI, the whole topic of AI managed second brain “LLM Wiki” came up, allowing to hoard even more information.

    And yet … I still do not use either of them.

    And even though it would be cool ro play around with the technology, my main questions before setting up one ofthe tools are:

    • Did I ever miss something?
    • Does it solfe an actual problem?

    And currently .. both are no for me.

    I just stumbled across Joan Westenberg’s article “5 Rules for a Second Brain You’ll Actually Use” with the more than accurate subtitle: “Tend a Garden, Don’t Fill a Warehouse”.

    And that just nails it so well for me: I do have my note taking (currently in Joplin) for all kinds of intersting stuff, plus the articles that I share here in the “Recommended Read” category. And .. I do not NEED more.

    I achieved to live a simple enough life so that I have no projects or so that are so complex that I need more organization than that. And I’m happy with it :-)

    Fediverse reactions

  • How to fix: VLC long startup time

    Since a while I noticed that when I double clicked an MP4 file, VLC took really long (like several seconds) to open and show the file. Actually really annoingly long. But as I don’t watch a lot of movies, I never really bothered debugging it.

    BUT recently I came across on itsfoss.com and a tweet on X from VideoLan themselfes that explained the problem and solution:

    The Problem:

    This is because of a « bug » from Microsoft Defender in one update of Windows 11, who magically made VLC plugins cache quarantined by Windows.
    Reinstalling VLC or regeneration of the plugin cache will fix this.

    VLC Team on X

    The solution(s)

    Clearing the cache by one of the following steps:

    • Regenerate the plugin cache: launch with vlc.exe --reset-plugins-cache
    • Clear VLC’s cache under the user profile at %APPDATA%/VLC
    • Reinstall VLC from the official build (download here)

    But this won’t prevent Defender to possibly kick in again.

    Therefore, excluding vlc.exe (exclude the process, not just the file) in Microsoft Defender should be a sustainable solution!

    The exact clickpath might change. But it should be under (german):

    Einstellungen >
    Viren und Bedrohungsschutz >
    Einstellungen für Viren & Bedrohungsschutz - Einstellungen verwalten >
    (ganz unten) Ausschlüße hinzufügen oder entfernen

    Fediverse reactions

  • Eight Myths on Software Engineering and GenAI – acm

    Actually, I just posted this on Mastodon, but I so agree to all those points that I want to reference them here on the blog as recommended reads as well.

    Generative AI is reshaping software engineering—but the narrative has gotten ahead of the evidence. Marketing claims, anecdotal wins, and misread studies have given rise to a set of persistent myths that are quietly driving poor decisions about AI adoption, tooling, and how to measure success.

    This article examines eight of the most common misconceptions.

    https://queue.acm.org/detail.cfm?id=3807963
    • Developers Spend Most of Their Time Writing Code
    • Writing Code Is the Bottleneck
    • Lines of Code Written by AI Is the Best Measure of Impact
    • AI Helps All Tasks and Engineers Equally
    • AI Will Turn Individual Developers into 10x Developers
    • It’s Up to Each Developer to Make AI Workgg
    • High-Performing AI Tools Will Be Adopted Automatically
    • With GenAI, Enterprises Can Innovate at Startup Speed

    See the full article here: https://queue.acm.org/detail.cfm?id=3807963

    Fediverse reactions

  • Backup your Crontab …

    Yesterday evening I just quickly wanted to change something in my crontab .. I was tired, I was distracted, I just SSH’ed into my machine and typed crontab -r.

    Ahh typo! I obviously wanted to crontab -e which is just ONE key away from -r. But … no error, no “unknown option”? This left me in a “eh what is -r and why is crontab -l empty?” – and I sensed a little disaster …

    A quick check with man crontab confirmed my discovery:

           -r     Removes the current crontab.

    Oh man. I never thought about backing up my crontab! Until yesterday. Now I have a fresh crontab and new line in in it, too:

    crontab -l > crontab.bak && git commit -am "auto commit" && git push

    Luckily there wasn’t TOO much in my crontab, but still, this was an anoying experience. Maybe there are smarter ways to backup scripts in my home directory – but it’s okay for now.

    Fediverse reactions

  • Debugging with Mistral: Learning, Not Just Fixing

    Depending on which social bubble one is into, people are drawn to letting LLMs do everything. Latest tools promise to fully configure or even manage entire server setups. While that sounds amazing (and I want to try it out some time) that approach has some downsides:

    1. Loss of control: I don’t really know what’s happening under the hood. If something breaks, I’m left in the dark.
    2. Loss of Learning: I want to know / learn how and why things (don’t) work.
    3. Digital Sovereignty: Relying on an external connection is creating a hard dependency that I want to avoid. And hosting my own LLM isn’t an option as my mini PC doesn’t have the RAM or power for that.

    So, while these tools are impressive, they’re not for me — at least not now or not in my private useacse of maintaining my main home-server.

    But, that’s no reason to not leverage it at all!

    (more…)
    Fediverse reactions

  • Local Models for Coding

    Coding agents are becomming more and more popular (well – I guess). Yet the dependece to LLM providers is pretty obvious: Starting in June, GithubCopilot changed to usage-based billing. In April and June Anthropic’s Fable 5 and Mythos 5 being available, then not, then limited, … And since then I read more and more about locally hosting models.

    To be honest: I was quite sceptical and surprised at the same time.

    Sceptical because the larger models were always told to perform ways better and surprised to see those extremes: The outcry of Fable 5 (large frontier model) not being available – combined with “let’s self host models” (meaning a fall back to WAAYS smaller models – so why this outcry of Fable not being available …).

    But then also reading articles about “Best open-weight models for coding” (09 July, 2026) … So … do we really NEED the big models? Is it just hype? And how much (V)RAM and GPU do I need?

    Luckily, Birgitta Böckeler did a comparison in two articles where she’s testing different models:

    Overall though, the agentic coding capabilities are definitely very far away from what I’ve now become used to with bigger models.

    Check out her articles. I would say they are worth reading to get an impression of what you could expect.

    Fediverse reactions