Results from the ASIC puzzle
Jane Street reveals the solution to its open-source ASIC puzzle, showing how hundreds of engineers reverse-engineered a chip designed to play Star Battle.
read it A technical reader will enjoy the breakdown of diverse reverse-engineering approaches, from custom C++ netlist extractors to SAT solvers and AI tools.
summary
Jane Street shares the solution and community write-ups for its August ASIC puzzle, where 400 participants reverse-engineered a GDS layout without a netlist. The chip was revealed to be a hardware checker for an 11x11 Star Battle puzzle designed with SKY130 open-source tools. Solvers used a variety of ingenious methods, including custom C++ extractors, Z3 and SAT solvers, dynamic impulse testing, and AI-assisted scripts. The post also highlights hidden easter eggs embedded in the layout and announces a new open-source protocol-emulator ASIC competition.
key points
- The mystery chip is a hardware validator for an 11x11 Star Battle puzzle, checking row, column, region constraints, and adjacency rules across 121 clock cycles.
- Participants used diverse methods like LVS netlist extraction in KLayout, custom evaluators, dynamic system identification, and automated SAT solvers.
- The layout featured numerous embedded easter eggs, including Morse code, a Jane Street logo, and an intentional floating-wire bug.
- Jane Street announced a follow-up open-source ASIC design competition focused on programmable protocol emulators, with submissions open until January 2027.
who should read it
Hardware engineers, reverse engineers, and participants interested in chip design, open-source EDA tools, and hardware puzzles.
discussion positive
consensus
Commenters debate the ethics and societal value of high-frequency trading funding such engaging technical outreach, alongside discussions on using LLMs and SMT/Z3 solvers for the challenge.
pushback
Some push back on viewing HFT as a purely wasteful enterprise, arguing that it improves market efficiency and lowers spreads.
I asked Claude to "fuzz the circuit" and, while it didn't work, it basically ended up going the Z3/SMT route and solving it in ~3 hours. This is after the writeups were already out, but still, as a software reverser, pretty wild. Writeup: https://www.babush.me/how-not-to-solve-jane-streets-asic-puz...
Provides a practical perspective from a software reverser who used LLMs and SMT solvers to tackle the circuit.
Generated by an LLM from the article and thread. Check the sources before relying on it.