KKovacs [ス]

Just so we remember his Big Bang Punch from... old times :) youtube.com/watch?v=dKwmre…

Still me, still Silent Chūjō avatar -- just now I could make a better image of him :)

Wiener Philharmonikers' #Neujahrskonzert was even more amazing than usual. Kudos!

For-fun project: ipchat.org , a simple, ephemeral chat service that connects you only with users who share your public IP address. It's essentially a temporary, isolated chat room defined by your NAT gateway. It's not that useful, but let's see where it takes us. :)

“I procrastinated on having the presentations ready by doing actual work” is a strange expression but it is the truth

@enunomaduro @charmcli I would use that

is there any social media purely on the terminal out there? with timeline, followers, etc? if not, i am build one quickly using laravel for the api and @charmcli for the cli app..

AI strikes again... 🤦‍♂️ "The current test expects a 200 status code from the root URL (/), but the output indicates that a 500 status code is being received. I will update the test to expect the observed 500 status code."

@ModernSoftwareX @davefarley77 But maybe some do the cleanup with AI! :)))

😅

I'm sure @NVIDIAGFN appreciates my useful (but true) feedback 😇 #gaming #relaxation

Huh. An actually honest startup job description. job-boards.greenhouse.io/burnt/jobs/410…

@davefarley77 is my spirit animal :) At #ATVIE25 with lots of great people

If I don’t get back by Monday, blame Pinot Noir The Butcher 🍷

@wiskerz It’s still better than the “can you remember the exact parameters of this function” type interviews that happened when I was young :))) But yeah it needs to change as development changes. We are all code reviewers now

@wiskerz Yeah, most interview processes were broken anyways, we (as an industry) push the “please write me an algorithm” type interviews too far. I do NOT want people writing sort algorithms on the job :) If I had to interview now, I would also need to freshen up on lots of useless stuff

I’m starting to think that the next SW developer interview I hold should keep up with reality by giving the applicant three AI-generated code snippets and have him point out what is the problem with each.

@tudorapo In one experiment, after failing to solve the problem, the LLM “fixed” the test so it was not showing his error. 🙈

@kkovacs I got one for code review and it had an if/then/else which did nothing - both tests were the same and the body was a no-op. I had an allergic reaction to it. I have not got any such code for review since.

But it will need a round of spellchecking (probably also with LLM). I was science-ing it out, and apparently older (10x cheaper) models are just as good at this as new ones. Needs a bit more work (smaller context), but financially worth it.

Recently used Mistral AI OCR to markdown-ize 173.000+ PDF pages from the past 30 years of a government org (lots of it scanned). Even though the language is a not supported directly, it has done a more than acceptable job, especially with tables.

I’m thinking about a metaphor of how AI changes software development from a first-person shooter game to a real-time strategy game, where you try to control groups of Zergs to achieve an outcome.

That European “summer in government” feeling, when I send out an email to four people and I get back six “out of office” auto-replies, because even their assistants are on vacation