Old Fucking Punk @lwriemen@librem.one

Indeed. As @olivia often says, guardrails are a scam. Model providers can play catch-up with stopgap measures but fundamentally these systems are open, and just as water seeks its own level, prompts find their way

Perhaps of interest, I go into some detail about guardrails and preference tuning (RLHF) in a recent paper of mine
https://doi.org/10.5281/zenodo.19452871

RE: https://infosec.exchange/@david_chisnall/117358990093685067

100%. The whole "terrorists can use an LLM to build a bomb" is a dumb movie plot threat. If you want to hurt people, lack of knowledge is not your problem. And nobody needs to develop a new chemical to do terrorism. All you need is readily available.

This is just "AI" labs creating imaginary issues to distract us from the real harm they cause.

I made some minor design and content structure tweaks to the Practical Accessibility course website (ahead of its third anniversary). https://practical-accessibility.today

One of the changes I made was creating a dedicated page for the many many reviews & testimonials that the course has received over the past 3+ years.

A heartfelt thank you to each & every one who has shared their experience & recommended the course. I’m forever grateful to you all. 🫶🏻

More to come!

“In response to “The death of web development education” - Andy Bell”

https://bell.bz/in-response-to-the-death-of-web-development-education/

> The most effective way I can contribute, in my opinion, is a crude sketch I just did of Piccalilli‘s Stripe chart for all time sales:

That graph is eye-opening.

RE: https://mastodon.social/@campuscodi/117353868814624656

meta being evil as usual

RE: https://mastodon.social/@UnicornRiot/117356196405647815

words fail at the sheer evil of #israel's actions here

Forgive me for posting a link to this site, it is important because some of you do not hate effective altruists enough!

Effective altruism is a dangerous cult: https://www.realtimetechpocalypse.com/p/effective-altruism-is-a-dangerous

#AI #noAI #technofascism #technooptimsm

Recently saw someone talk about their prestigious AI specialisation to a group of veterinarians, who then walked out thinking that the guy was an expert in Artificial Insemination and asked me "So what's ChatGPT?"

I keep getting the feeling that integrity is a luxury the obscenely wealthy cannot afford.

Filled out an online form, and to get a copy of my answers I am required to sign into Microsoft. Hell, no. One more reason I despise the greedy control freaks who run Big Tech at this point.

“The Tech Industry: An Autopsy | Computational Impacts”

https://monroelab.com/2026/09/29/the-tech-industry-an-autopsy/

> The ‘AI’ scourge, is more than a propaganda campaign, it has mutated (as all things do) into a sort of religion. Unsurprisingly, considering the combination of capitalist mania, resurrected eugenics, and hatred for workers that spins, like a molten core, at its center, the scourge is relentless, deforming everything it touches.

Look out in the coming weeks for a trailer and official release date for the 2nd film of our full-length documentary series, InterRebellium.

Hey #selfhosting adjacent people, reminder that with two commands on #Debian you can have a self-hosted #Tor Onion service running:

sudo apt-get install --install-suggests python-txtorcon
twistd web --port "onion:80" --path ~/public_html

The above will make your "~/public_html" (yes that's how old I am) available on a brand-new Onion domain.

More details:

https://txtorcon.readthedocs.io/en/latest/introduction.html#introduction

We noticed that the database was slower to respond to queries the last few days.

This was caused by a specific query that is executed for quota calculation. This query took longer than normal (bad query plan) to get a result. MariaDB spent most of its CPU time on those queries.

We instructed MariaDB to re-analyze the relevant tables; this resulted in MariaDB using a better query plan and giving the expected performance for the query.

This unfortunately caused some not planned downtime.

RE: https://mastodon.social/@markgurman/117354990486859177

So layoffs and price increases, the same as every tech company.

If #AI can be wrong on this, it can be wrong on the correct internal temperature for safe to consume chicken.

It might give you the right answer, but it might give you the wrong answer. And I have to wonder - yes this is a little conspiratorial - if that isn't the point.

If you cannot uneducate a population, perhaps you can flood it with information that is so deeply unreliable as to make any fact at all tenuous and ephemeral. 7/

RE: https://hachyderm.io/@dalias/117354507865860424

Stop believing the narrative of despair they want us to take in from "1313 CVEs!" and dig into the actual data, or look for reports from others who have. It's a huge lie.

@siddhesh_p @jschauma They could at least outsource the triage to the reporter.

If they can't even be bothered to explain why they think the bug is a vuln, it shouldn't get a CVE assignment.

If their explanation is technobabble out of the slop extruder, it also shouldn't get an assignment unless someone else is able to determine it warrants one.