This article is part of a series of articles that discusses possible advancements of weaponry. The Episode 0 focused mainly on ammunition and propulsion systems of the ammunition. This article, the Episode 1, is mostly about bombs. (Yes, imagine that: a blog post about "how to make a bomb", uuuuuuuuuu ).
The short answer is that explosive based bombs might not be such a good idea, regardless of the purpose of the bomb. There exist chemicals that are so reactive that their VAPOURS burn through multiple layers of rubber gloves. The dispersal mechanism of those chemicals might be based on the compressed air or electric pumps. A chemical resistant "Coca-Cola" bottle that is under considerable overpressure and is opened by burning through a cable tie that locks the lid, might be pretty effective. The burning of the rope can be controlled with electronics. The electronics can be something truly simplistic, for example, a light sensor with an amplifier, RC-toy-motor controller or a plain circuit from a mechanical switch and a battery. Probably there are many ways, how to construct the trigger mechanism without any electronics, but the electronics gives the weapon more "precision".
It is known that neural networks can be stored to a relatively small amount of memory. If a neural network is trained to trigger on the voice of a particular person(read: car type, transportation vehicle engine), then that neural network can be used for a bomb trigger that probabilistically blows up only, if that particular person "talks" "long enough". If the longevity requirement did not exist, then the scenario might be that Mr. Hitler comes home from work, puts his Nazy coat away and shouts, "Honey, I'm home!", KABOOOM. (Hitler and chlor seem to fit well together.)