Peter Favrholdt retweeted
This is Genie 4, our most realistic and interactive world model yet. 🌐
Explore dynamic environments in real time to see how the scene responds, sounds, and how the people in it move, react, and speak.
Peter Favrholdt retweeted
So basically, today PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse, and it works on WordPress and MediaWiki and Magento and Joomla and Laravel, and Censys says like 1.5 million servers are running a version with no fix, so the NSA was like "take your system offline," and the FBI was like "take your system offline," and CISA and the UK and CERT-EU were all like "take your system offline," and then the Prime Minister of Japan, the actual Prime Minister, posted at 5:55 in the morning telling 48.8 million people to take their system offline, so everyone took their system offline, like the Dutch cybersecurity agency took its own website offline, and the Dutch parliament went offline, and the Port of Rotterdam went offline, and Finland and France and Denmark started pulling government sites, and Wikipedia went down, and Archive.org went down, and Asana and GOG and Something Awful went down, and the University of Vienna's webmail went down for some reason, and that's just PHP. Because GitLab also dropped a CVSS 10.0, which is the highest score they give out, for a bug that's been there since 2014, so that's every self-hosted GitLab for twelve years, and then LiteLLM 1.107.0 got hijacked and shipped a credential stealer and a remote access trojan in every single wheel and the Docker images, including latest, and you didn't even have to import it, just installing it was enough, and it's the second time this year, and then Railway paused every deployment on the entire platform because someone got into the environment variables, so now everyone has to rotate every secret they've ever stored there, and TiDB Cloud said someone was walking around inside with stolen credentials for basically a month and an extortion group says they took 10.6 terabytes, and Zeabur got breached for the second time in two months and is just shutting down, like the whole company, and Node.js patched remote code execution and a TLS bypass, and CISA says Plex is being actively exploited, and someone took over 300 Tor exit relays to steal SSH passwords, and SafePay ransomware has hit over 600 critical infrastructure orgs, and India's biggest bank filed a data loss incident, and Japan has been in an official state of emergency in cyberspace since October 9, which honestly explains the 5:55 AM post. Oh, and eMudhra, a certificate authority that's issued over 60 million certificates, whose entire job is proving websites are who they say they are, was supposed to be using private keys that are random numbers over a hundred digits long, and one of them was 4. Just 4. And the others were like 66 and 237. Anyway, it's Wednesday.
I går prøvede jeg Mimir, der er en ny dansk open source-sprogmodel, udviklet bl.a. på SDU. En af min kones kolleger har været med på projektet, så jeg besluttede mig for at prøve den lokalt på min RTX 5090.
Her er link til pressemeddelelsen fra SDU:
sdu.dk/da/om-sdu/institutter…
Idéen er god: hvis man ikke kan være med i kapløbet om at bygge den bedste største model, kan man jo forsøge at bygge den bedste mindste :-) Mimir har kun omkring 1 mia. parametre og er ifølge pressemeddelelsen optimeret til at få meget ud af de få parametre.
Mimir er ikke tilgængelig direkte i Ollama endnu. Jeg fik efter lidt bøvl bygget en patched llama.cpp og derefter koblet Mimir op på OpenWebUI.
Og så fandt jeg ellers ud af at context window kun er på 4096 tokens! Det er for lidt til at være brugbart i praksis og noget jeg kunne have læst mig til inden jeg gik i gang og sparet bøvlet :-)
En ny "tom" prompt i openwebui er allerede på 5428 tokens. Det overskrider Mimers context window og dermed er modellen ubrugelig til længere chats, RAG, dokumenter og agent-workflows.
Men selve idéen om en lille dansk effektiv model synes jeg er interessant og håber de en dag også kommer med en model med større context window.
Måske kan den begrænsede Mimir bruges som 2. trin "korrekturlæser" på det halvdårlige dansk, der kommer ud af Qwen3.8:27b? (det kræver lidt ekstra udvikling, så nu må vi se om jeg får lavet det).
Verdens bedste mindste model? sdu.dk/da/om-sdu/institutter…
Pretty amazed by qwen3.8 running locally on RTX 5090 using ollama. Built an NDI viewer on Debian and figured out solving the missing ffmpeg7 libs all by itself - really useful!
Peter Favrholdt retweeted
The anti-carnivore drumbeat is getting a little predictable.
Someone starts carnivore and feels dramatically better. Inflammation improves. Energy stabilizes. Cravings disappear. Weight comes off. Metabolic markers improve.
Months later, after they’ve built a healthier foundation, they experiment with adding back a little fruit, vegetables, or other whole-food carbohydrates.
Cue the critics:
“See? It was never a long-term solution.”
I see it differently.
For many people, the elimination phase is the point. It removes the noise, gives the body a chance to recover, and helps identify which foods actually work for them. Once they’re in a healthier place, some discover they can tolerate foods that previously caused problems.
That’s not failure. That’s progress.
Some people thrive on strict carnivore for years. Others settle into an animal-based or low-carb approach. The goal was never to earn a dietary label.
The goal is better health.
Find the approach that delivers the best results with the fewest downsides—for you.
N=1 beats diet ideology every time.
Peter Favrholdt retweeted
Obviously ppl should carry a CO2 monitor at all times :) Outside air is ~400ppm, stuffy room ~1000+. CO2 ppm is proxy for how much other people's air you're breathing (~covid risk). Thinking gets hazier at 1000+. Meeting rooms and bedrooms can climb much higher than you'd expect.
Peter Favrholdt retweeted
🚨New Paper: "Seven Years of 700 Cholesterol Without Coronary Atherosclerosis: A Lean Mass Hyper-Responder Case Report"
Link: doi.org/10.3390/diseases1405…
For the past 7 years, I’ve been running what is essentially a natural experiment in cholesterol and heart health.
During that time, I’ve largely lived with:
👉Total cholesterol around 700 mg/dl
👉LDL cholesterol between 500–600 mg/dL
I recently underwent advanced coronary CT angiography imaging with AI-guided analysis. This is not a CAC. It measures all plaque (soft + calcified), with expert interpretation and AI-guided analysis capable of quantifying plaque down to the cubic millimeter (mm3).
Now, to address the obvious question:
Am I too young for plaque?
In brief: No.
The clearest comparison is individuals with homozygous familial hypercholesterolemia, who often have similarly extreme LDL/ApoB levels and can develop advanced plaque as toddlers, and even heart attacks as early as age 8.
Also, nutrition influencers in their 30s have publicly shared quantified plaque scores from these same imaging technologies. In one recent case, a plant-based influencer in his thirties was found to have 61.3 mm³ of plaque despite having far lower lifetime LDL exposure. (He can identify himself if he so chooses.)
My case also isn’t a one-off.
There are many individuals like me, including older individuals with similar LDL-C and ApoB without any plaque.
The difference is that I’m an unusually well-characterized subject, with extensive metabolic data and health markers tracked over time. You can learn more at the newsletter or open-access paper, linked above.
The science of heart health is not settled. And cholesterol is not a simple story.
🚨 If you want to help spread the word...
Quote Tweet this post (or create an original post) including the article link with a thought. Academic papers are increasingly evaluated using attention metrics. Original posts from unique users are one way to increase these metrics and help ultimately increase its reach.
🚨 If you want to learn more, I'll include more learning resources below 👇
Peter Favrholdt retweeted
CO2.
CO2!
CO2!!!
Low CO2 is all plus, plus, plus.
For her PhD research, Dr Lara Tookey found that when classrooms had fresh air and low CO₂ levels, students performed better on cognitive tests.
airspothealth.myshopify.com/…
Peter Favrholdt retweeted
20 years ago, after a licensing controversy, the linux kernel project lost the free access to its version control system
existing alternatives at the time simply couldn’t meet the project's requirements
so Linus Torvalds set out to create a new system - one that is distributed, fast, and reliable
a system that would go on to become the de facto standard, adopted by nearly 95% of all developers today
this is the story of git
Peter Favrholdt retweeted
Ilya Sutskever (OpenAI cofounder):
"What does it mean to predict the next token well enough? It means that you understand the underlying reality that led to the creation of that token"
Peter Favrholdt retweeted
🚨🚨🚨BREAKING🚨🚨🚨
#LMHRstudy / KETO-CTA published!
100 lean, metabolically healthy #keto:
🔥 No association of #ApoB with #plaque ↗️
🔥 No association of #LDL #cholesterol with #plaque ↗️
🫀What is association? Existing plaque
Read and share 👇👇👇
jacc.org/doi/10.1016/j.jacad…
Peter Favrholdt retweeted
In 1975 this garage builder’s kit changed the world. Prior to its release only “qualified” folks could touch a computer.
These first micro computers caused “experts” and elite to laugh.
Yet they inspired the very device you are using.
The same is taking place with AI today.
Peter Favrholdt retweeted
"Al offentlig dansk IT bør fremadrettet baseres på open source-software – både til udvikling og drift. Det er kun gennem open source, vi sikrer fuld uafhængighed og suverænitet over vores digitale infrastruktur." world.hey.com/david/hvorfor-…


















