if it's obvious to everyone that verification is the bottleneck to AI accelerated software engineering,
why are so few people learning math? not linear algebra or analysis, but stuff directly related to verification, like proof, type, category, lattice theories?
I cant stand these millennial tech companies with these 'cute' nonsense names like "granola" or "clay." if I had a company youd see me calling that shit the American Computing Corporation. National Electron. United Tabulation. thats a name with some chest hair
@uscompco has some strategies for mitigating this for our hardware. Software badly needs a lot more visibility and analysis tools, there's SO much low-hanging fruit.
In the long term, @uscompco is the computronium company.
And true computronium is a full-stack problem - math, physics, manufacturing, chip architecture, software, and back to math.
@scheminglunatic Out-of-order negates almost all of this advantage in modern CPUs. The trail of red dots in the bottom right corner of the graph is a mix of x86 and ARM chips.
Sure P and NP are cool, but do you know about coNP? What about the Σ and Π towers?
Are you aware of the Polynomial Hierarchy?
Do you know the difference between PH and PSPACE?
Are you familiar with descriptive complexity? How these classes all map to First Order Logic with different numbers of quantifiers?
Have you glimpsed a fragment of God?