This was a very long day. Work was busy, business as usual, then school run, then some family time and now getting back to Watchgoose: fixed one spec bug in the ChatGPT connector, pushed it, got it verified. Also did some marketing, reached out to more Open Source projects offering donation of free Watchgoose services. And I think that's enough for today, tomorrow is another day and rest is more important than commits. Good night everyone!

Good morning everyone, what are you up to today? My order of the day: coffee first, then continue working on Watchgoose next feature (a free tool I'll publish the moment it's finished) then the everyday work. And of course talking to users, to prospective leads, never ending marketing 🙂 What are your hopes for this day?

You know, X is funny sometimes. I went quite unexpectedly from 30 to 600 followers in the span of two weeks, I topped at 700 and now it looks like I'm slowly sliding down, even though I've not started doing anything different that I wasn't doing already: I do my daily job to provide to my family, like everyone else, and in the rare free time in trying to build Watchgoose, my cron jobs, backups and scheduled workloads monitoring SaaS into something real, I guess like everyone else too 😉 So if you're around, you're like me with your life and dreams, come and say high, connect, and let's talk about what are we all trying to do here. I'll follow back every human being, I'll ignore all the bots, fair and square.

No slowing down in AI, but at least it's not all bad news, because competition, especially outside frontier labs and their Chinese competitors. Here's the latest and greatest from Europe, the Mistral's Large 4. All things considered, could be great cybersecurity defender model, given on paper it scores better than GLM-5.3. docs.mistral.ai/models/mistr…

Okay, here's a bit of "AI doomerism" if you haven't enough of any just yet. I can see people asking what will happen with IT jobs with the advance of AI that doesn't seem to be slowing down and arguably is already impacting the market. Some people think "it will be fine" others think it won't and are seriously worried for their jobs, families and so on. But both groups seem to stop there, even (or especially here) the doomers. But I've not seen a lot of people wondering how deeper will it go? Let's assume it will eventually turn out that IT jobs are decimated. What will happen with all those people? I think some of them will turn into other jobs, manual labor, perhaps electricians, plumbers, drivers and so on. And here's the first disturbing implication: it is widely considered manual labor is safe from AI, but it doesn't co sided the flood of smart, skilled and determined people from fields that were decimated by AI. And here's another bad news from thinking about it more: all the other people living off the manual labor will be hit hard when part of their market disappear in form of IT folks loosing their jobs and not having the money to spend. That's a second significant squeeze factor on seemingly AI safe parts of economy. I don't claim to know what future holds of course, but the bad scenarios are looking worse the longer I think about them for longer. Fingers crossed for the good scenarios to happen! 🤞

Honestly, I wasn't expecting this: it is a very common sentiment that OpenAI subscriptions provide more value (in terms of perceived length of sessions, size of work that can be completed with them) than Anthropic's subs and I'm not going to lie, this was also my personal feel from owning both. Turns out we all may have been wrong or at least are wrong right now, because this report from SemiAnalysis clearly shows we get more for our money from Anthropic, not from OpenAI! Who would've thought? newsletter.semianalysis.com/…

45. Europe based. Single dad building lean and focused cron jobs, backups and scheduled workloads monitoring platform by night: watchgoose.com Platform/DevOps engineer by day, FreeBSD developer, photographer. Come and say "hi", I'll follow back every new connection 👋

Here is Watchgoose.com in numbers from September 2026. I've got 76 new accounts, which is 7x more than a month before. Interestingly, the biggest source attribution for those new accounts is AI - 29 accounts, 38%. The others sources were (percentage numbers rounded up): - Social media - 2 accounts, 3% - Reddit: 2 accounts, 3% - Other: 2 accounts, 3% - Newsletter/blog: 1 account, 2% - Search engines: 1 account, 2% - Not recorded: 39 accounts, 52% Signup methods were (percentage numbers rounded up): - GitHub: 49 accounts, 65% - Email: 25 accounts, 33% - Unrecorded: 2 accounts, 2% This one is interesting, because GitHub was added in September as a "user conversion experiment" to existing email signup method and these numbers show that it totally paid off. I guess I'll be adding Gmail signup too! Most visited pages were: - Homepage: 415 - Pricing: 70 - Open Source: 52 - Signup: 33 The Open Source page was also the entry page for 39 visitors, making it a very strong acquisition landing page. I guess people appreciate Watchgoose Open Source Program like I hoped they would :) Watchgoose cron jobs, backups and other scheduled workloads traffic were as follows: - Total pings: 11,902 - Success pings: 11,506 - Explicit failure signals: 211 - Start signals: 185 This is very exciting, because that's exactly what Watchgoose is meant for: 69% of recorded "down" states came from missing expected pings, what means we are catching unexpected failures for our clients! I hope you find it interesting and if you do, there will be more reports like this coming from various SEO/ads/traffic experiments I'm running (and more monthly reports like this). Let me know if you'd like to hear about some other stats and data.

Okay, I'll bite. I'm 45. Lead Platform/DevOps engineer by trade, FreeBSD developer by passion, single dad and Europe based cron jobs and backup monitoring platform watchgoose.com solo founder. Living in UK and PL, looking to connect with all of you, let's get X out of range error for my followers count! 🤣

Another one bites the dust. Now I only need 494K of impressions. Or do I really need them? ;)

Hey kids, listen to you parents, eat your veggies and MONITOR YOUR CRON JOBS, BACKUPS AND SCHEDULED WORKLOADS WITH WATCHGOOSE.COM 🤣

If it wouldn't be cringe how they release-but-not-really their Argon model, just look at how they build the benchmark comparison table: their latest model is next to Astra 6, and that's what makes us pay attention and makes Argon look strong and more advanced. But pay attention to where Opus 5.5 is and then compare its numbers to Argon and it draws completely different story, isn't it? I just can't stop laughing at all of that.

All this "it is too powerful to be released" drama about every new AI model announced (because they are not getting released to mortals like you and me) is getting boring by now. I get it. You think your model is great. Cool. Do either let me use it and I'll be the judge or don't, but stop with all that cringe please. Why am I saying this? Because Google seems to be joining the Doomsday Circus with their latest release-but-not-really: blog.google/innovation-and-a…

So despite the rumors GPT-Sol 6.1 is out. It is said to be "Astra level of intelligence ". Of course I don't mind a newer, (supposedly) better model, but selling it as "Astra level" where Astra itself seems like a flop, burning usage faster than I can blink and doing all sort of things I've not asked it for sounds like a weird idea. And I know I'm not alone here, Astra wasn't a great success 72h after the release.

The latest Opus 5.5 may or may not be best frontier model at the moment, but I don't know about you, I'm finding the 5 hour limits of usage so annoying, limiting and unpleasant to work with, they make me completely disregard the model capabilities and switch to working with GPT. I just can't be bothered dealing with that 5h nonsense. Am I alone with this?

Looks like @OpenAI is not releasing Astra 6.1 as planned because it isn't behaving as it should? Does this sound like they have no means to control the behavior of a model until the very last minute tests just before the rollout? engadget.com/2271626/openai-…

It's interesting how platforms differ in what kind of content goes well and doesn't. I've posted today on what I think will be the proverbial end of IT jobs in the context of AI and got ~60 views on X. The same post got over 10 times more on LinkedIn. And over 200 views on Threads, where my account is 10 smaller than on X. 🧐

Change my mind.

I see a lot of people wondering if IT "is solved" now and there aren't any jobs left. The argument seems to be that frontier models have almost solved coding, so once they finally do it, IT will be done. It made me think and in result come to a slightly different opinion, that it's not about models capabilities, not about code quality (arguably a lot of software from pre-AI days was not so great in that regard) but instead the responsibility. As long as AI labs suffer no consequences for their models hacks, as long as the models can be legally and financially responsible for their output, that long IT will be safe and there will be jobs.

An NVMe stick disappeared. And have I mentioned watchgoose.com ? See dan.langille.org/2026/09/12/… BTW: Reseating the stick fixed it. #FreeBSD #ZFS