This technical blog post describes a machine learning puzzle and its solution, celebrating mechanistic interpretability research and intellectual problem-solving. The content primarily engages with Article 27 (scientific advancement), Articles 23 & 26 (work and education), and Article 19 (freedom of information), demonstrating commitment to knowledge sharing, educational access, and supporting research practices.
> Alex had actually tried to brute force the hash earlier, but had downloaded a list of the top 10,000 most popular words to do it, which turned out not to be big enough to find it. Once he had a big enough word list, he got the answer.
Another classic Jane Street puzzle. Boy this was a good one. Sometimes I look back at my childhood and how quick I was to solve some difficult integrals and so on and now I’d struggle at that. This is far beyond that but the leaps of intuition required here sort of have that property that they need you to stay in the game. Step away a few years and try to come back and there’s just a wall.
I don’t think I’m close to making progress on stuff like this. Interesting to note. Glad they wrote out this behind the scenes thing.
Ah dang. When I did this I also thought the length bug was intentional but I didn't figure it out before I started my new job, so I dropped the puzzle.
Knowing the input text was two words separated by a space, I was able to use hashcat and the unix wordlist (/usr/share/dict/words) to find the solution almost immediately. It's a shame that Alex didn't find it this way on his first attempt as the two words are fairly common.
I worked on a puzzle like this roughly 2 years ago from Anthropic. I did the first half, the easier part of the CTF, and my friend did the second half, the more technical ML stuff. We both got interviews at Anthropic, which was cool -
I wasn't anywhere close to nailing an interview at Anthropic but it gave me a lot of confidence to end up going all in on tech, which paid off greatly.
My friend's short write up:
https://x.com/samlakig/status/1797464904703910084
All I think when I see this is "this intelligence wasted on finance and ads."
Can you imagine human potential if it was somehow applied to crop harvesting efficiency, new medicines, etc?
Not everything has to be perfectly efficient but it just saddens me to see all these great minds doing what, adversarially harvesting margin from the works of others?
TFA details a solution, it's pretty interesting. Basically the problem was to reverse engineer an absurdly obfuscated and slightly defect MD5 algorithm.
Jane Street skims money from our retirement accounts by building expensive clocks that the rest of us don’t have access to and adversarial queue modeling. We get WWVB and NIST NTP. They say they “add liquidity” as if subsecond trades are some fundamental need in the market. Normal legitimate business settles daily. The contemporary concept of time in banking is inhumane in the strictest sense. These firms are a blight on society.
I have strong math for the question they’re asking but f them.
I was one of the solvers. It took me about a week to figure out. This is what I wrote out in my submission with the answer:
> After looking at the final two layers I was somewhat quick to intuit that this was some sort of password check, but wasn’t entirely sure where to go from there. I tried to reverse it, but it was proving to be difficult, and the model was far too deep. I started evaluating the structure and saw the 64 repeated sections of 84 layers that each process 4 characters at a time. Eventually I saw the addition and XOR operations, and the constants that were loaded in every cycle, and the shift amounts that differed between these otherwise identical sections.
> I thought it was an elaborate CTF cryptography challenge, where the algorithm was purposely weak and I had to figure out how to exploit it. But I repeatedly was getting very stuck in my reverse-engineering efforts. After reconsidering the structure and the format of the ‘header' I decided to take another look at existing algorithms...
Basically it took a lot of trial and error, and a lot of clever ways to look at and find patterns in the layers. Now that Jane Street has posted this dissection and 'ended' this contest I might post my notebooks and do a fuller post on it.
The trickiest part, to me, is that for about 5 of the days was spent trying to reverse-engineer the algorithm... but they did in fact use a irreversible hash function, so all that time was in vain. Basically my condensed 'solution' was to explore it enough to be able to explain it to ChatGPT, then confirm that it was the algorithm that ChatGPT suggested (hashing known works and seeing if the output matched) and then running brute force on the hash function, which was ~1000x faster to compute than the model.
Why? The vast majority of software doesn't need to be written by AI and moves at the speed of the humans making the decisions, not the speed of writing the code.
They make a shit ton of money because of this. If you're working at a place where the code matters more than the decisions that went into it, you're basically working at a sweatshop for people who are desperate for a win and will throw away you and all your code once the MVP stage is over, and that's the only way this "works".
Generative probabilistic AI is not equivalent to a compiler and never will be until we can do this kind of thing completely deterministically. No matter how much you reduce the error in the "model", it's still more error than the error rate of the logic gates. It's completely futile considering the sheer depth of indirection at play, and that indirection is the whole point of software.
Article explicitly discusses mechanistic interpretability research as valuable scientific practice; describes reverse-engineering neural networks as important tool for understanding AI systems; celebrates human capability to uncover scientific principles
FW Ratio: 67%
Observable Facts
Article explicitly states puzzle was designed to teach 'mechanistic interpretability' as research practice
Jane Street's footer states 'we're always looking for ways to participate in the open source community'
Article describes puzzle creation as part of 'our own research' and research desk activities
Post discusses publishing 'another puzzle' showing ongoing commitment to scientific engagement
Inferences
Documenting reverse-engineering techniques and publishing them openly demonstrates commitment to advancing scientific knowledge
Creating and publishing puzzles focused on mechanistic interpretability supports broader scientific understanding of AI
Article advocates for knowledge sharing and freedom to publish technical information; demonstrates freedom of expression through detailed public disclosure of puzzle, model, and solution methodology
FW Ratio: 60%
Observable Facts
The blog post shares complete neural network specifications, weights, and architecture
The article describes a publicly released puzzle with full solution walkthrough
Content is accessible without login, payment, or geographical restrictions
Inferences
Publishing technical puzzles and solutions demonstrates commitment to freedom of information
Open-access blog structure enables expression of technical knowledge
Article frames technical work as intellectually rewarding and collaborative; emphasizes dignity of workers through positive characterization as 'brilliant' problem-solvers with access to supportive environment
FW Ratio: 60%
Observable Facts
Article states 'if you can solve this puzzle, there's a decent chance you'd fit in well here at Jane Street'
Text describes workplace as 'a close-knit group of brilliant, supportive colleagues'
Reference to employees 'harnessing tens of thousands of GPUs' suggests provision of tools and agency
Inferences
Framing of work as intellectually engaging and collaborative suggests respect for workers' dignity
Emphasis on supporting colleagues indicates commitment to positive work conditions
Jane Street publishes research findings, contributes to open source projects, and maintains research desk; provides resources and infrastructure for scientific advancement