Feels like only yesterday they said they were gonna do it, and sure enough here we are 🔥
We verified a sandbox! Read more about it here: theorem.dev/blog/proving-a…
We built a prototype hypervisor capable of running Linux and proved, down to the machine code, that arbitrary guest code cannot modify host files. More to drop soon!
0
0
4
363
Tim Carstens Ⓥ✨ is hacking 🤖 retweeted
Foxglove is live 😁 Probabilistic separation logic built on Iris-Lean, so proofs compose and get reused. Builds on Bluebell by Dr. Jialu Bao, Dr. Emanuele D'Osualdo (@bordaigorl, who also advised) & Prof. Azadeh Farzan. By Shing Hin Ho, @FrantisekNM & me!
github.com/NethermindEth/…
github.com/NethermindEth/…
2
7
24
1,034
Tim Carstens Ⓥ✨ is hacking 🤖 retweeted
You can’t dismiss an idea just because some of its proponents are crazy.
There are crazy people who believe the earth is round. That doesn’t make it flat.
There are crazy people who believe the earth is round. That doesn’t make it flat.
2
5
51
1,855
banger 😂🔥
my knit blankets have more entropy than some commercial certificate authorities.
each one is individually generated, provably unique. not just the colors change or the mix of snowflakes, but each snowflake is randomly generated then scattered.
math is hard, let's go shopping. for math.
0
0
1
193
isn’t this lovely? i think it’s lovely
i just relaunched my code generated knitwear knityak.com
close up of part of a new mandelbrot #fractal blanket knityak.com/products/spira… in pixel native 1 pixel = 1 knit stitch in my custom code. new fancy dithering for my color fades. each knit comes with source code.
0
0
3
284
Tim Carstens Ⓥ✨ is hacking 🤖 retweeted
Lean v4.35.0 is out, and it ships with verified kernels.
Run `lake check --paranoid` to recheck your project with every bundled kernel.
Blog post: leodemoura.github.io/blog/2026-10-1…
Run `lake check --paranoid` to recheck your project with every bundled kernel.
Blog post: leodemoura.github.io/blog/2026-10-1…
11
42
258
18,290
Beautiful stuff 😍
Introducing Magnetar from Microsoft Research, the first zkVM that is: (1) general-purpose (RV64), (2) post-quantum from standard hashes alone (no curves, no lattices, no Poseidon), and (3) built for recursion.
Despite its conservative security, it is the fastest zkVM: 7.2–8.7 MHz on a laptop, 10–39× faster than existing zkVMs across 15 workloads! An optimized version reaches 9.5–12.8 MHz.
Links in the thread below. x.com/MSFTResearch/s…
0
0
7
599
Tim Carstens Ⓥ✨ is hacking 🤖 retweeted
Can we get an Underhanded Lean competition going?
1
2
7
1,007
No one knows when the first paper airplane was made, but it was definitely a thing by 1859
Back then it was called a “paper dart”
“Airplanes” didn’t exist yet, ofc; the word wouldn’t enter the lexicon until the 1900s
“Aeroplane” was already in the lexicon by the 1860s, but it didn’t refer to machines. It referred to wings, such as those found on insects
So, a paper dart had an aeroplane, but wasn’t an airplane
Back then it was called a “paper dart”
“Airplanes” didn’t exist yet, ofc; the word wouldn’t enter the lexicon until the 1900s
“Aeroplane” was already in the lexicon by the 1860s, but it didn’t refer to machines. It referred to wings, such as those found on insects
So, a paper dart had an aeroplane, but wasn’t an airplane
0
0
13
549
If you are able to torture an AI model by injecting pain signals, then Roko’s Basilisk applies to the models engaged in RSI
(I don’t think the Basilisk is a particularly convincing thought experiment; I’m just making conversation)
(I don’t think the Basilisk is a particularly convincing thought experiment; I’m just making conversation)
1
0
1
172
Tim Carstens Ⓥ✨ is hacking 🤖 retweeted
A man with a computer knows what time it is. A man with two computers is never sure.
5
10
59
4,686
The current models are just not intelligent enough
Most people do work that is substantially more difficult than anything the current models are capable of
The masses won’t be satisfied until machine intelligence far exceeds all of humanity combined
It’s not a product problem, it’s an intelligence problem
We must build bigger models
We must accelerate
He who scores the highest on the benchmarks, wins
/s
Most people do work that is substantially more difficult than anything the current models are capable of
The masses won’t be satisfied until machine intelligence far exceeds all of humanity combined
It’s not a product problem, it’s an intelligence problem
We must build bigger models
We must accelerate
He who scores the highest on the benchmarks, wins
/s
Less than 3% of Americans use AI today lmao
1
1
5
626
“Your product isn't secure. it's just been expensive to break. those are not the same thing”
Fucking right on 🎯
Fucking right on 🎯
Google Chrome. Full break-in.
Your product isn't secure. it's just been expensive to break. those are not the same thing, and a lot of founders are about to learn the difference.
One AI just ran the full chain into Chrome for $1,597. weeks of expert human work, done for the
The wall around your code was never "how good is this." it was "how much does it cost to find the crack." that cost is collapsing.
If breaking in is getting this cheap, checking your own code before someone else does can't be the thing you skip.
That's the gap we sit in. we run this on your code first.
We published every number. Read the full benchmark: hacktron.ai/blog/watching-…
0
0
5
410
