Used grok for fuzz debug as it was once the pareto model with least cyber refusals before deepseek updated its flash.
Looking forward for this iteration
At SpaceX, we write our own headers and calculate our own time (includes leap seconds). Not many people will know what this means, but those that do will appreciate it.
The new Grok AI/SI training and inference stack is derived from incredible of work done by amazing engineers to create Starlink.
1
131
Solar for fuzzers: endgame
AAAND it's a win! 19GBs plateaued
1. Echidna: vibed memory optimizations rolled back to original echidna memory management og @0x310f1sh + some more optimizations
2. Solar: two bug detected, one is fixed - not in release yet, second is smaller but harder, reported. So currently compile solar from main
Solc - 2h 04 mins compilation
Solar - 4.8 sec
Solar for fuzzers pt. 3
Remarkably heavy suite, same echidna version:
- Solar 4.8 secs compilation, RAM goes >103 GBs
- SOLC 2h 04 mins compilation, RAM 13 GBs
Solar for fuzzers: @DaniPopes hinted Echidna might have a hardcoded way to read the solc info, and after a couple of quick fixes we got pretty coverage now (still to be triple-checked/verified - all vibed)
Compilation time 75 mins -> 7 secs
Btw works only for 0.8.0+ suites
1
6
412
Solar for fuzzers: currently a no
1. Currently echidna crashed on Set.elemAt after the first call
2. Coverage: echidna can't connect source to the opcodes solar is producing = lines with no instruction = no info in coverage files
3
8
572
Indeed
Replying to @natashenka @5aelo
Overall, fuzzing and LLMs are complementary approaches. AI has greatly increased the number of vulnerabilities found and fixed in V8, but don’t turn off those fuzzers yet!
1
283
AI this AI that. Run the fuzzer and find the x1000 cases of protocol DOS in no time
🤖 Made with AI
1
209
FORKFORK DOG retweeted
9/ where it breaks down: proofs are blind where they abstract. I planted a bug in 4 lines the spec summarizes away — writes routed to the wrong order book. every proof stayed green. unit tests caught it instead.
so no, you don't get to delete your tests.
continuing. one solution would be to create a cgroup like `cgroup memory.max=7G` and retry
3 workers and sitting on 7GBs
1
1
61














