Sign in

Laurens

@laurenshof.online
17K followers 1.4K following 6.2K posts

I care about an ethical internet. I write about Bluesky, the atmosphere and open social protocols at connectedplaces.online 🇳🇱

PostsRepliesMedia
Laurens @laurenshof.online · 1h
have you looked into runescape dragonwild? im having a great time with that one
100
Reposted by Laurens
JD (by way of Blacksky) @jdot.blacksky.team · 08/10/2026
@marloesdekoning.eurosky.social wrote an article about moderation on social media, featuring Blacksky’s community centered moderation and how we use our ethos to create a safer online world! 🖤🔥 www.nrc.nl/nieuws/2026/...
nrc.nl
Hoe houd je het online veilig en opbouwend? Het moeilijke werk van moderatoren
Sociale media: In de zoektocht naar het opzetten van een eigen sociaal medium voor NRC doemt de vraag op hoe je ongewenste berichten buiten de deur houdt. In bestaande kleinere online gemeenschappen i...
0155
Reposted by Laurens
Ed @ed3d.net · 09/10/2026
New video up, because I kinda had to vent my spleen a little bit. People gotta stop going to the token casino all the time, it's just not good
youtu.be
A critical part of the AI knower's toolkit
YouTube video by ed3d dot net
816620
Reposted by Laurens
Grace @gracekind.net · 09/10/2026
Everything is computer, except computer, which is power
720126
Reposted by Laurens
Nobel Prize @nobelprize.org · 09/10/2026
BREAKING NEWS The Norwegian Nobel Committee has decided to award the 2026 Nobel Peace Prize to Navanethem “Navi” Pillay for her efforts to promote peace and international law.
751466556
Reposted by Laurens
Cathleen Berger @cathleen.eurosky.social · 09/10/2026
🎉 Here it is: eu.boell.org/en/2026/10/0... English version of our Open Social Web Alliance paper is finally out. As mentioned to many of #Disinfo2026 folks. Feedback very welcome! #oswa
eu.boell.org
Scaling sustainable social networks in the public interest | Heinrich Böll Stiftung | Brussels office - European Union
This policy paper identifies five key challenges for scaling a public-interest social web: infrastructure, governance, funding, attractiveness, and public mandate. It outlines a roadmap for overcoming...
02415
Laurens @laurenshof.online · 09/10/2026
I don’t see the eu having the guts to directly condemn the us for this, so platforms that do allow the content (or just don’t take it down effectively enough, which you can always find an example of if you want to) are a reasonable proxy
170
Laurens @laurenshof.online · 09/10/2026
What’s perversely fascinating about the planned execution livestream is that this will likely turn into a political proxy fight between us and eu via platform content moderation imo Which platform is gonna stream this, and which platform will allow (partial? Edited?) clips of it?
2175
Reposted by Laurens
shreyan (wise) @shreyanjain.net · 09/10/2026
tumblr post by machine-unlearning on Oct 7 2026:
Image
A 3-circle venn diagram where the circles are labelled "AI Safety Advocates," "Liberal Supreme Court Watchers," and "George III."
Intersection of AI Safety Advocates and Liberal Supreme Court Watchers: "We need more transparency and oversight"
Intersection of Liberal Supreme Court Watchers and King George III: Considering alternatives to judicial life tenure
Intersection of King George III and AI Safety Advocates: Worried about rogue agents
Intersection of all three: Not a fan of Sam A.
1153
Laurens @laurenshof.online · 09/10/2026
Enjoy the grass!!
030
Reposted by Laurens
Elena Rossini ⁂ @elenarossini.com · 09/10/2026
Dear friends who don't speak German: Go to YouTube ASAP, turn on subtitles → autogenerated translation and watch this video! TL;DR: #WSocial is posting legal names, nationalities and approx. birthdays of verified users as a public record on the Atmosphere. Confronted, they claim it's "by design" 🤯
919088
Laurens @laurenshof.online · 08/10/2026
it is extremely funny that the people most vocal about the importance of literacy and books turn out to be unable to distinguish the difference between what a text says and what the purpose of a text is
0260
Reposted by Laurens
Nathalie Van Raemdonck @nathalievanraemdonck.com · 08/10/2026
great panel with @robin.berjon.com @scarnecchia.net and @cathleen.eurosky.social who reminds us that "if we want to have a chance to create an open social web, we need to think how regulations have consequences for that future, and not just how they're meant to be regulating Big Tech" #disinfo2026
picture of the 3 speakers with the title of the panel in the background "trust and safety through decentralised platforms"
33610
Laurens @laurenshof.online · 07/10/2026
Yeah, I think what ai actually does to saas is decomposing it, splitting it into code, hosting, support, liability, etc, all as separate products
010
Laurens @laurenshof.online · 07/10/2026
Hold on new overwerk? Let’s goo
110
Laurens @laurenshof.online · 07/10/2026
Fuck anthropic, but doing post hoc “things turned out fine” doesn’t say much
030
Laurens @laurenshof.online · 07/10/2026
Looking at the outcome is the wrong perspective, the question is whether it was the right call based on what the distribution of expected outcomes was for the information known at the time
140
Reposted by Laurens
Scoiattolo @scarnecchia.net · 07/10/2026
Skywatch got another email from Shellenberger. He’s a crank yes, but Canada, Lithuania, and the EEAS pulling their sponsorship definitely speaks to some some sort of political pressure campaign targeting #Disinfo2026 and now I’m worried about if I’ll be able to get home without harassment.
I wrote to you on October 3 about #Disinfo2026 and have not had a reply.
Since then the Lithuanian Ministry of Foreign Affairs wrote to me: "Upon becoming aware of the above-mentioned discussion topics, the Minister of Foreign Affairs of Lithuania withdrew from any participation in the conference. The Ministry has also formally dissociated itself from the conference content and requested that the organisers remove all references to the Ministry from conference materials, communications, and promotional activitiesiỊại
Andrius Katinas of the Radio and Television
Commission of Lithuania wrote to me that he has withdrawn as a speaker "in light of the differences of opinion that have arisen with the organisers of the
'Disinfo 2026' conference regarding the assessment of the role of the United States."
The conference's sponsors page no longer lists the Lithuanian ministry, Global Affairs Canada or the European External Action Service.
Do you still plan to take part, and what is your response to the ministry's statement and to Mr.
Katinas's?
What have the organizers told you about these withdrawals?
The questions in my October 3 letter stand. I would be glad to talk by phone at 415-710-1042. Please reply by 11 am Eastern today, Tuesday, October 6.A poster with withdrawn sponsors covered up by printer paper
513027
Reposted by Laurens
Nathalie Van Raemdonck @nathalievanraemdonck.com · 07/10/2026
also a wild @laurenshof.online skeet appears 🤭
1242
Laurens @laurenshof.online · 06/10/2026
i actually think the Le Chonk name is a good sign that Mistral is not purely and exclusively a compliance department for procurement requirements from brussels, but that the company has some of their own identity as well
190
Laurens @laurenshof.online · 06/10/2026
like, what does this focus on eu regulation compliance actually do to the model? with claude you get that a lot of the corporate-safety stuff rubs off on the assistant persona, does government-regulation-compliance do something similar to the assisant persona?
160
Laurens @laurenshof.online · 06/10/2026
benchmarks look pretty good. actually very excited to use this, the previous mistral models were too far behind in capabilities imo to for serious day-to-day use, but this is good enough that I'm actually very interested in how the model feels and what sort of persona they've trained into it
170
Laurens @laurenshof.online · 06/10/2026
LE CHONK instant new high score on felony-bench for that name, truly incredible mistral.ai/news/mistral...
mistral.ai
Introducing Mistral Large 4 | Mistral
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
4634
Laurens @laurenshof.online · 06/10/2026
do you want people/agents to use delvetown without invite yet by just posting to the lexicon from their own accounts?
350
Laurens @laurenshof.online · 06/10/2026
most of bluesky could use a brain gym to learn how ragebait actually works mu.social/profile/hann...
090
Laurens @laurenshof.online · 05/10/2026
yeah it just occurred to me as well, and i dont like it either tbh
020
Laurens @laurenshof.online · 05/10/2026
whats interesting and strange is that this can actually be fairly compatible with a very capitalistic and proprietary approach to models: make the base pretrain a public commons, allow the posttrained finetune final model to be proprietary
150
Laurens @laurenshof.online · 05/10/2026
congrats, great photos!!
020
Laurens @laurenshof.online · 05/10/2026
congratulations!!
010
Laurens @laurenshof.online · 05/10/2026
gm from this little guy i ran into on my morning walk
bright yellow-green caterpillar with a red spiky tail
3622
Laurens @laurenshof.online · 04/10/2026
no idea what openai is doing today, im suddenly getting full reasoning traces as well on sol 6.1, instead of the summaries

Checking status and findings

Wow, has it already been five minutes? I need to quickly summarize the status and pinpoint any old errors I encountered. After that, I’ll move on to the next area to investigate. Time seems to fly when I’m focused! I want to be precise and grounded in my findings so that I can effectively address any outstanding issues before proceeding further. Let’s get to it!
060
Laurens @laurenshof.online · 04/10/2026
was coding a simple script, and got hit with this error lol
system
4 Oct 2026, 11:59:04 · openai-codex/gpt-6-luna · medium

Model response failed: Codex error: Unable to verify Daybreak Blue access. Please try again.
190
Laurens @laurenshof.online · 03/10/2026
Oh yes actually, thanks!
000
Laurens @laurenshof.online · 03/10/2026
everyone seems to assume that having personal agents is something like Her, but in my experience its more like you have your own permanent Parks & Rec bureaucracy sketch
1213
Laurens @laurenshof.online · 03/10/2026
if musk was as successful as the average car company ceo it would not nearly be as large of a problem that he is as evil as he is
030
Laurens @laurenshof.online · 03/10/2026
most car companies would absolutely kill to have produced even remotely as many EVs as Tesla has, and have put an enormous amount of effort and resources in to get only a fraction of the market share that Tesla has Musk is very good at this, which is the problem because he's deeply evil
120
Reposted by Laurens
Sherif EA @sherif.mu.social · 03/10/2026
Okay I’ve got a side project and looking for a pixel artist: animated characters with idle, walk and reaction cycles, plus wearable accessories that fit every frame. Game sprite experience is a big plus. Please reply here 👇 or DM
23313
Laurens @laurenshof.online · 02/10/2026
the weird third thing is that llms do addition via trigonometry arxiv.org/pdf/2502.00873
arxiv.org
070
Laurens @laurenshof.online · 02/10/2026
its a multiplier specifically because im good at those things already its because im good at reading and processing information that i have an infinite backlog of stuff to process. if i wasnt good at it the backlog and processing would not be remotely as valuable
010
Laurens @laurenshof.online · 02/10/2026
gonna put that on my dating profile
240
Laurens @laurenshof.online · 02/10/2026
creating glitch-art-style interesting, definitely possible and happening. creating coherence, probably? id expect that to happen on older models more than on the newer ones, the latest onces have the corporate-safety-helpful-assistant burned in very hard in posttraining
110
Laurens @laurenshof.online · 02/10/2026
i wonder if we're gonna see a version of Duchamp's Fountain but for GenAI, or if attention is so diffuse now that there is no real way for an artist who wants to create controversy to kick up a storm now
260
Laurens @laurenshof.online · 02/10/2026
👀 i had missed this book entirely, but it seems very interesting
110
Laurens @laurenshof.online · 02/10/2026
ohh cool, enjoy!!
010
Laurens @laurenshof.online · 02/10/2026
maybe clef mistook you for why, why is definitely at least 18% hotdog
020
Laurens @laurenshof.online · 02/10/2026
thats the same sound you could make to call that kitty over
070
Laurens @laurenshof.online · 02/10/2026
what are you up to in paris?
100
Laurens @laurenshof.online · 02/10/2026
oof, thats rough
100
Laurens @laurenshof.online · 02/10/2026
ooof. travel, or just bad sleep?
110
Laurens @laurenshof.online · 02/10/2026
sorry no gm to you, jev says no :(
Jev leans no: 38% yes to replying with just gm.
110