Someone who invested their life savings in dot-coms in 1999 is probably never again going to invest in "the next big thing". That brings its own form of irrationality - once everyone believes that it's foolish to invest in the next big thing, the next big thing becomes very underpriced - but they are at least not going to make the same mistake twice.
Beyond the fact that you would need a direct link between genes and rationality for selection pressure to apply, you're positing that rational people have more reproductive success. Is that really true you think?
I was contending with the overly broad criticism of the government (and other institutions) as protective figure through laws, rules and regulations. Letting the system evolve without rules by exploiting irrationality, letting individuals learn exclusively by mistake is a bad mechanism. His "evolutionary" argument could apply to most rule systems -- do you believe rules are inherently bad? That's nonsense imo.
The assessing of mistakes and teaching is done by domain experts (rule makers or law makers), who know much better than new entrants the perils of a given system, in the case of a well functioning system. There is also evolution in those systems, but it's a more effective form: usually when a single actor finds an exception or flaw in the rules, the rule makers are warned and may rewrite, propagating this knowledge to everyone.
If we had infinite capacity to learn, reason and absorb minutiae of the world around us, we wouldn't need rules. But if you go, for example, work in a dangerous factory environment, you don't want to study about all the systems and every way you could die, or even worse (per the 'naive evolutionary' suggestion), try yourself going everywhere. You want someone to tell you "You can't step outside those lines; when working at this machine, you can't put your hand in those places", because you trust the rule maker has gone through everything and you won't walk out missing a limb. Is this "babysitting"? Probably, it's still the best way to go about it.
That's not to say every activity should have strict rules. Rules are good fit in general where the environmental conditions don't change very rapidly (in time or across the actors) and where the perils they protect against are sufficient to warrant learning and enforcing the rules in the first place. They remove flexibility for each actor to make it's own optimum choice (only relevant when this varies in time or across individuals) and take resources to formulate, teach and enforce.
If you're being intellectually honest here, you need to go the full distance and acknowledge that your logic (if actually scientific) requires the "irrational" genes to NOT be passed on (at least not as much as the "rational" ones).
So... eugenics or death.
Considering they can sneak in CISA into a US budget bill after it was shot down by the public, apparently, neither does the government.