My fav thing in coding agents are agentic benchmarks. I even don't prompt. And the output goes where it belongs i.e., to /dev/null ;-)

Anthropic published a silly block diagram and everyone is like "Anthropic just changed agentic workflow OMG".

WTF is going on.

It's a shame that I live in the first floo, cannot even jump of the balcony ;-)

I end this with Claude alike prose.

"It used to be only three blocks.

Now there are many."

🧛

@monsieuricon I wrote a longer take elsewhere, which is IMHO somewhat balanced point of view :-) Or at least I feel that I've found myself the right balance how I see the world from here forward ...

"I think the strategical challenge for companies and workplaces is actually when not to use AI i.e., balancing between human labour and automation.

ATM, the balance is completely broken, which is not shocking because new tech trends cause some mayhem and chaos.

My experiences on this topic are:

1. Linux kernel. I just write patches by handle like always. The screenshot shows what happens if I try to use AI with Linux kernel. When you know one project really well, it might be also easier to just write code. Otherwise it's like "C-to-English-to-C-WTF-#%%##%-C-AGAIN!?-..." generator, and the only outcome is what can be seen in the screenshot.

2. When using codegen for anything, have at least a single unplugged day in a week. It really can keep slop under control, and it helps to keep your AI use in those four other days simple, effective and productive :-)

3. Despite having all that knowledge and deduction even the strongest frontier model is just a fancy non-modifiable database with a hardcoded program run by inference at the data center. I strongly believe that the real benefits come only when a really good lawyer uses AI for "law stuff", developer for "developer stuff" and a scientist "sciency stuff".

Maximalists, shitfluencers and that type of scum and scammers have tried to sold the world the idea of human labour being obsolete and only an expense but really it's the only fuel that makes thing actually work.

One for the humankind, cheers 🍻"

Powered by Buildroot :-)

#buildroot

I created a proof-of-concept fork of pi-hashline-readmap plugin, integrating sandbox to its compressed bash output pipeline:

https://github.com/jarkkojs/pi-hashline-readmap

It's a bit more non-trivial integration example because bash hook cannot just be repealed and replaced.

With a very thin and unintrusive layer that Landlock LSM provides and SECCOMP notifications further help to make more robust Landstrip provides very lean integration path for sandboxing coding agents in Linux. And yeah, stuff like below can be described with this.

While working on my appliance OS build etc. I've let Hindsight to profile the sessions. There's one common theme and it is that in no time I'm in all caps :-) It's often that I write the code in other tmux pane with vim while complaining to prompt in other. So yeah, this is how long-term memories look like ATM.

With kernel code, it's a suicide mission. I don't get the "AI assisted" part in those. It's IMHO the hard way when it comes to Linux. Tried it couple of times and I'm happy that my laptop is still in one piece...

And yeah, whatever they say, I'm not sure how these long term memories are going to help :-)

There's now partial network sandboxing with a combination of Landlock rules and seccomp notify.

The irony is that I don't even have Claude Code installed. I had an account early Spring, which I asked my employer to remove later during Spring because, other reasons aside, I don't fully trust the company providing this great service. If something is way below the market price, that alone should rise suspicions, or at least it does for me.

As per sandbox, I just made the logical conclusion that it is best to pick the widest spread. And also make Landstrip "repeal and replace fit" to Anthropic's sandbox is one of the goals.

2027 is the year when there's no text to scan and further it is already feedback loop. Scanning images and videos is sort of like multiplying everything collected and required compute with an astronomical number. That year is like the expiration date of frontier LLMs in a way.

This can't be real. I'd recommend to delete all your forks at Github.

It $0.29 AI bill because of a fork.

The reason I've been making now so much AI noise is the realiziation that

1. I have bad vibes only ;-)
2. It is probably better to start taking baby steps right now with sec.
3. Got postulated that guardrails for malicious stochastic actions should be based on algorithm, not AI.

There's a lot of popular lore of some kind of guardian LLMs that overwatch frontier model but the problem is the introduction infinite recursion of distrust. All LLMs have the same underlying problem.

Explained to LinkedIn what mathematicians do as a profession as apparently this was not clear in the first place :-)

#mathematics #openai #linkedin

found some random unidentifed sun glasses while cleaning up my place. now all i is a van, duct tape...

Still useful and also frequently used references in 2026 :-)

#x86 #arm #assembly

@caesarcattus I otherwise agree but would exclude FSF despite they do a lot of that too but it is of transparent and aligned with what they say they do.

I've never really grabbed e.g., purpose of Open Source Initiative. It's basically a branding agency making anything that comes from SIlicon Valley appear to be aligned with open source.

I would have really hard time to find a person who would willingly try out Goose. I could almost bet my money that people who picked it have never tried it :-) Screenshot this time really speaks for itself.

This generative AI is so amazing, everyone gets the exact same web site.

😅

spotted from linkedin

Vim 9.2 :----) let's go...

The lightline theme called (drumroll) monochrome is the only theme I've ever made.

There it is: the AI monster.

Grace Blackwell is the cultural icon of this decade delivering the slop.

has been a long time, at least two years, but making some music with bitwig. let's see what this turns into :-) just early drafting...

I like spreadsheets. They are a great help for figuring out stuff. For Google's services I use exactly the spreadsheet, Keep and YouTube, and that's it.

I've been wanting a TUI spreadsheet for a while but I only recently found sc-im. This is superb :-)

I need to figure out at some point how can I replicate something like GOOGLEFINANCE with Lua scripting capabilitie. I use it all the time in Google's spreadsheet.

I'm developing my own ostree and BuildRoot based operatingy system:

- https://codeberg.org/puu
- There will be puu-desktop and puu-llm builds.
- puu-desktop uses ext4.
- puu-llm uses ext4 + xfs for the models.
- puu-llm hosts k3s, llm-d and generally very latest of stack for running local LLM modesl.
- GNOME build is tough and I'm working on it but I have plan how to make in happen on BuildRoot.
- Given being BR2_EXTERNAL, builds and configures with beloved kbuild :-)

Despite having puu-llm build this project disregards AI contributions.

That said. I don't see it as a crime that you generate some initial code with LLM.

I do that sometimes too especially when I need to get an idea of e.g. alien projec. However, usually I do also spend week or few applying hard manual labor to make the actual code change I want to do despite having some random slop that functionality fits the bill.

This is what I do accept but it is generally best just to say plain no in README.md as most people will likely not be able to do such weighting.

[The logo image artwork is copyrighted by me (drawn with Inkscape) and licensed with CA BY-SA 4.0]