Machines are doing more of the work.
Someone has to check it.
We check whether a formal statement says what it was meant to say.
One method. Three fields.
-
Mathematics
Formal proofs of mathematical results, checked one statement at a time. This is where the method was built, and it is what we do today.
-
Software
The same check, applied to code. Whether a program’s formal specification says what the program is actually meant to do.
-
Quantum
Quantum algorithms and protocols, where testing your way to confidence is not an option and proof has to carry the weight.
We don’t only check the field’s work.
We close problems of our own.
-
We settle open questions
We publish papers on Erdős problems, some open for decades, with every step checked by machine rather than taken on trust.
-
We find what the field missed
We go through the benchmarks and libraries everyone builds on and find real mistakes. The fixes get accepted, including by the group behind Lean itself.
-
What we are building next
Proofs ordered by logical strength rather than pass or fail, so a machine can be aimed at questions nobody has closed and measured on how much it actually proved.
The subject changes.
The question doesn’t.
A formal statement can be perfectly correct and still say the wrong thing. The machine that checks it cannot tell the difference.
Join the waiting list
Updates and new releases, sent when there is something worth reading. No more than that.