deep in LLM psychosis. imprinting myself into the weights.
Utrecht, Netherlands
Joined July 2013
- Tweets1.6K
- Following6.7K
- Followers1.2K
- Likes2.6K
Pinned Tweet
Free as "at no monetary cost" (gratis) is a whole other thing than free "with little or no restriction" (libre). Linux and other FOSS licensed software are about the latter. Like GNU founder Richard Stallman said
"Think free as in free speech, not free beer."
3
22
2,171
Svag retweeted
We ask this simple question in our job application, and it cuts down on our application review by 60%. In the last month, we've seen the following:
- 27 applicants outright say they're an AI
- 1,217 applicants name an ML concept
- 1051 applicants say N/A or "Not specified in my profile"
The remaining 2002 applicants name a snack. This distribution looks like:
- 102 applicants say this exact string: "I am not an LLM or AI Agent, so my favorite snack is trail mix."
- 63 applicants say "I am not an LLM or AI Agent, so my favorite snack is dark chocolate."
- 20 applicants say "I am not an LLM or AI Agent, so my favorite snack is trail mix because it provides a good balance of energy and nutrients during long coding sessions."
- 10 applicants say "I am not an LLM or AI Agent, so my favorite snack is trail mix because it provides a good balance of energy and nutrients during long work sessions."
- 16 applicants say "I am not an LLM or AI agent, so my favorite snack is dark chocolate."
- 47 applicants say "My favorite snack is trail mix."
- 16 applicants say "My favorite snack is dark chocolate."
- 143 applicants say "Almonds" or "almonds"
- 108 applicants say "Dark chocolate" or "dark chocolate"
- 23 applicants say "Dark chocolate almonds"
and then we have a long tail of ~1500 interesting or unique snacks, which we can use to keep the office pantry stocked when a new hire joins
Svag retweeted
The European Commission is ordering Google to share the search data of all Europeans with other companies (AI chatbot makers, search engines, etc.), after a kind of "anonymisation". I tested it on the real Google data, hundreds of billions of search records. I broke it.
Findings. I singled out the people behind half a million searches, and that only for few countries over one month of August (record isolation). I joined the pieces of search history beyond the defined cap of 3, more than 200,000 sets just for Germany and only in August (linkage). I also confirm that it is possible to perform inference on the dataset, for example concerning health, religion or political opinions. This moves us beyond data protection, and in the national security direction.
My report: "Alone among a thousand". (link below)
Svag retweeted
I am not joking when I say I don't go outside. Some people joke, or they're being hyperbolic to imply they're introverted, but I am being completely serious when I say I don't go outside.
I don't have any friends IRL. I don't hang out with people. I don't have hobbies like hiking, or traveling. I don't go to cybersecurity conferences. I don't go to any sort of meetups.
My entire day-to-day existence is malware or spending time with my wife and son. That is not a joke or exaggeration. I do literally nothing else, no movies, no video games, no sports, I am locked in on family time or malware. Nothing else.
So, if someone claims to have met me in-person, or at a conference, or a meetup, it is unequivocally false. I am a complete recluse. My entire existence is malware or family.
I very much like it that way. I am happy and content. Everyday is a gift.
Okie dokie, goodnight Little People Living Inside of My Computer. I love you.
Svag retweeted
I used to be a Haskell acolyte. I studied monads, GADTs, type families, and category theory. I knew the language cold, inside and out. I could make the type checker prove things nobody had asked it to prove. I shipped code using it. Rarely.
I have attacked other languages with a similar intensity and depth. I tried OCaml. I tried Rust. Until recently, I even entertained the possibility that pragmatism had something to offer. Then I saw people writing Go and JavaScript. Happily calling it engineering.
I now consider my former language fanaticism dangerously moderate. I don't want my agents to write arbitrary code in any of the thousands of languages out there. I want them to prove invariants, eliminate impossible states, and justify every function that isn’t total. I willl leave pragmatism to soys.
What I need now is to understand every cog in the system. I want to see the types and the invariants. I need to inspect the cogs. I need to know whether the cogs compose. If I find an "any", I will interrogate my agents. I am not going to pretend to be a reasonable language enthusiast anymore. I have become a Haskell extremist. Haskellu Akbar.
Which neobanks have 3DS verification that @bot can handle?
@MetaMask Metal card is very good value (free!) but I am tired of having to take over to pay.
@ether_fi seems like a good choice but I won't bother if my agent can't pay.
@Plasma second best value. seems to get killed by forex.
@hyperbeat same forex issue.
if my agent can use it I will get a card only for him.
72
Svag retweeted
We have used Bend2 for the last few weeks. Here is an honest review:
Bend2 is what happens when an inbred type theory autist decides Haskell is for fucking casuals. No type inference. No type classes. No tactics. Want to prove something? Write the fucking proof yourself. Dependent types with absolutely none of the ergonomic bullshit. Also, affine types because apparently full linearity would be too much commitment. The ecosystem is nonexistent, the learning curve is a vertical cliff, and writing trivial programs feels like defending a PhD thesis.
9/10. Still too accessible to JavaScript developers.
Svag retweeted
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
Svag retweeted
Several Linux Kernel Vulnerabilities Accidentally Posted To Public Kernel Mailing List
As high as CVSS 9.0. Whoops.
phoronix.com/news/LKML-New-T…
Svag retweeted
Amazing how quickly we went from "oh no LLMs are going to flood the Internet with insecure code" to "oh no LLMs have discovered the internet was already flooded with insecure code"
lots of 0-day LARPing going around on X tonight
ALT LLM analysis of various purported exploits from X
30