fock

I was on the train when some executive support staff joined my car (train ran late and they were easy to find on the internet ...). They behaved like misogynistic ogres and I can vividly imagine those people laugh about this. 0 regard for other people or their societal responsibility.

at the banking place I work running things in k8s, z/OS-people are actually the ones running custom git clients in go on z/OS. Bonus: they have no nosy Java devs (recurringly producing threading bugs...) saying "but we all use spring boot!!!" and likely no manager asking "is this cloud-ready???"

all my family paid happily for Whatsapp. Then we got refunded because Facebook bought them. Now we get ads. Yay :)

that might indeed be the problem. I however did not find any other suitably marked element and CTRL+F apparently is too simple for outlook. Maybe you can select a single mail in the search bar...

and then most of the places I know happily allow employees admin access for "just that piece of software they need" and simultaneously push for "zero-trust". There's no point in it at all and you can just as well use saltstack to rollout apparmor-policies on your locked-down linux (and suddenly the same people wanting GPOs tell you that linux is untenable because of usage restrictions)

yesterday: I saw the weird CVE for M365 which "exploits" some LLM through messaging embedded in emails.

today: got a very long email, wanted to search for our department in it. Outlook: "Search is a deprecated feature".

Despite all the "but you can't extrapolate to a large org from personal experiences"-FUD around, I think for most orgs (especially governments which are generally far behind on processes) it would be easy to switch from a feature-perspective. The problem is the army of employees and contractors who are very happy to defend Microsoft for keeping their non-automated thiefdoms (such is non-cloud AD administration at most places I extrapolate ....). There is hardly anyone there to implement the necessary processes and rather than to send out their underlings to FOSDEM, leadership is happy to get an invite by Microsoft (or a cloud-provider...) to an "innovation-summit" instead.

And so what? Given the display of men's feelings w.r.t. to "mixed groups", I (heterosexual male) get the ick about some people here... For me most of it is about the space/relationship where certain things should happen, but I guess scientific misogyny is a thing too.

also software is a factory and the LLM is a workshop building factories. Also I strongly believe that people building factories still do a lot of a) ten people pounding out the metal AND b) measuring to check

it's not sinking in the bureaucracy. There's an industry of outright fraud an embezzlement...

oh, they are available to startups. Startups having the sole purpose of skimming funds by being the technology partner to some academic.

That's the best case. Then there is outright fraud:

https://cordis.europa.eu/project/id/101092295 - European dynamic provides some project management and a wordpress-page for the lump sum of 800k€ and of course there is always "SOCIAL OPEN AND INCLUSIVE INNOVATION ASTIKI MI KERDOSKOPIKI ETAIREIA" headquartered here: https://inclusinn.com/. Probably still in stealth mode, using the 4M€ to "promote innovation".

So for workflows it's like Airflow, Brigade or hatchet or ...? How do workflows integrate with k8s (ressources, ...)? Camunda can also deploy natively on k8s. However you still develop apps for Camunda and it seems like dapr is no different there? Why is it in CNCF if it doesn't provide a way to build a workflow out of k8s-native artifacts (PVs, Deployments, Jobs, ...)?

While I don't play Badminton (and so can't test with a racket on hand) this seems very cool! I also thought about something similar for judging bikewheel spoke tension - I guess I have to research this a bit more now.

As for monetization: I personally don't have problems with static ads served from your domain. Find some celebrity or brand and ask them if they want to have you serve their banner.

Thank you, very nice read! (Though from some scanning it looks like it mostly helps reads)

I can tell you the company I work at (4000 people, legacy banking IT) has 4 people running our Datalake. We likely have more people buying/"evaluating" Databricks currently (from overhearing calls in open-plan offices), so I guess they have a point. A very sad point...

So how is this distributed Postgres still an ACID-compliant database? If you allow multiple nodes to query the same data this likely is just Trino/an OLAP-tool using Postgres syntax? Or did they rebuild Postgres and not upstream anything?

as others have mentioned

- institutional inertia - some weird consultant style people in key roles (this happens around cloudy stuff too) - the DBA-team - "we can't move everything!" - "we just migrated off solaris!"

however every new project with sane leadership seems to decide against oracle.

we pay millions to Oracle. We hit a bug and it took 6months for them to reproduce and acknowledge there is a bug. they now seem to be on the lookout for someone being able to produce a fix: sales and indian after-sales can't do that... curious!

Oracle seems just a moneygrabbing shell company at this point and I suppose the whole hyperscaler-cloud is developing towards that point with the leaders of those corporations repeating exactly the same talking points...

should the matrix-multiplication at the core of this not be in a core library? Why are generic layers intermixed with LLM-specific kernels when the generic layers are duplicating functionality in torch?

Upstreaming that might actually help researchers doing new stuff vs. the narrow demographic of people speeding LLMs on MI300X's.

I think they are taken over by exactly the same people leading the AI-hype. Funny how in this article they are a) not advertising clearly what they are doing, b) solving a small subset of problems in a way noone asked for (I think most people just want ROCm to work at all...) and c) just adding to a complex product without any consideration of actually integrating with its environment.

I guess it's vibecoding "AI"...

also why is it calling .cuda() to move tensors to a cuda driver? I suppose this is because this is based on HIP - which comes with it's own set of problems, but that's ROCm for the masses I guess.

Also the tgemm.mm has to be a torch module (at first I thought this was some lowlevel library which they now have a preview of, because there is a ROCm-torch already ...) which is evident from the table just before the summary. That table also smells like they are mostly focused on inference...

EDIT: seems official ROCm-torch is also based on HIP.

yes, automation is necessary and from what I see most non-IT companies are far behind what's possible. Something which is not necessary and imho pretty stupid is replacing idiotic workflow components designed to wash-out responsibility with LLMs and now let the bots run amok (which probably is totally ok with the bureaucrats who designed these workflows in the first place). Most of the AI automation crowd supposedly thinks this is a high value activity - while looking down on the bureaucrats. Which is pretty funny to me...

the author might enjoy acquiring a pressure washer from his boss! Decisive action to take back his personal liberty and getting a proper high-pressure cleanse will be beneficial to humanity for sure.

For reasons I recently got acquainted with enterprise scheduling solutions. IBM/HCL workload automation specifically - the company is actively looking for a replacement but nothing except for the chinese sponsored open source comes close it seems to me (and that is a step too far for management...).

The thing you miss with scripting is that a good workflow engine essentially provides a simple and well-defined layer allowing you to couple many concurrent processes and providing ressource management. Curiously out of the box all the opensource workflow tools do so too - like you said people (Airflow has pools, but thrn ...) essentially built very weird DSLs to do high-level composition and there you can indeed rather easily code it yourself. I assume the companies originating these have integrations for the other bits too while everyone else does fake elastic scaling with some halfbaked workarounds. And from time to time someone comes up with a new DSL and some GUI - ignoring all the hard parts.

state-owned company where the 2 families owning the majority of shares block using the earnings to do any investment. The only thing were the state-stake comes into play is in not crushing unions in Germany (which they don't mind doing in the US) and not letting the whole enterprise fail (which it should ...)

just mount the whole root fs as a volume? or use bwrap or crun or runc to run your OCI bundle. Sadly the latter tools have some warts for this purpose...

going by the post you reply to it is code again.

huh? DBUS is very much a thing and has CLI-tooling?

Hmmm, k8s is OpenSource and IBM has rolled up a nice bundle to sell that. Did you think of the 80ies maybe?

ended up at a place like this too + in our project we, the devops-people, also develop and support the z/OS-data-integrations. I generally tell non-IT people it's like a sysadmin, but not the windows kind they interact with - you won't roll out fixes in 2 hours from testing to prod (including the process stuff) at that scale (PB of data-warehouse, some 1000 B2B-customers on different versions) if all you know is clicking buttons in Windows server.

as a non US-person who recently wanted to look at some cases: is there a free way to do this? Or do you have to sign up to the weird ePACER?