Imagine if you can randomly jumble up code and still have a reasonable outcome. DNA is not only an expressive language its end-product is rather durable to changes.
For me it's less expressive than assembler!
Probably the answer you are looking for is:
1) long code, each minimal part is longer than what you expected. Do you know how much code is necessary for a cell to decompose the sugar (and oxygen) into CO2 and water?
2) lot of superoptimization https://en.wikipedia.org/wiki/Superoptimization i.e. sometimes after a lot and lot and lot of generations it can by lucky and find a witty combination that does something. (the trick is not killing all the intermediate generations, so it's not a 100% superoptimization)
3) nice interface with the real word. The DNA is translated to RNA that is translated to proteins. The proteins are made of aminoacids. There are a bunch of aminoacid types, so they can build the proteins like modules. Some aminoacids are more water friendly, some more acid, some can make bridges with other aminoacids or other molecules. So it's like a nice construction set where you can pick some structural parts, some moving parts and some "sticky" parts. (A bad analogy is a Lego with magnets and velcros.) If you combine them wisely, you can do interesting things.
> Imagine if you can randomly jumble up code and still have a reasonable outcome.
If you randomly jumble your DNA you die. If you modify a single base of all your copies of your DNA you may die (some bases are important, some not, better not try). For example to split sugar you must make a protein that holds it. To hold it, it must have the correct shape. If the change in the base replaces a small aminoacid by a big one, the sugar will not lock in the hole of the protein and the sugar will not get split and you die. If the change in the base repace an acid aminoacid by a basic aminoacid, the sugar will not get stick inside the hole of the protein and the sugar will not get split and you die. If the change in the base repace a aminoacid that likes water by an aminoacid that hates water, the protein may not get the correct shape and the hole will exist, and the sugar will not get split and you die. If you are lucky, the change in the base will not cause a change in the aminoacid, so you don't die, but it's a risky experiment. If you are extremely lucky the chane will make all the process .01% more fast or efficient and your descendants may conquer the world. But the default outcome is die, die, die.
If you get a bunch of about to get pregnant mouses and irradiate them, you can get a lot of mutations in the eggs/sperm. But the problem is that almost all mutations are irrelevant or deadly, so with enough radiation you don't get weird baby mouses, you only get dead baby mouses (if they survive enough to be born).
If you randomly jumble your DNA only a little you die.
---
If you are interested, I recommend reading a biology book like the Curtis-Barner https://www.google.com/search?q=curtis+barnes+biology It's a big book and will give more context than you need for this question, but it's a good book.
Ray's Tierra (early 1990s) was a fascinating experiment with this. Its DNA are little assembler-like programs telling the 'creatures' how to reproduce, how to copy their genes. Many kinds of parasites evolved. I believe no-one had tried randomly mutating genetic info like that before Tierra.
https://en.wikipedia.org/wiki/Tierra_(computer_simulation)
http://life.ou.edu/pubs/doc/index.html#What
I recommend Sapolsky's Biology of Human Behavior course - accessible to beginners, it goes into a lot of detail about DNA, genes, the brain, emotions, sexual selection, evolution, and the many other levels involved in animal and human behaviour.
https://www.youtube.com/watch?v=NNnIGh9g6fA&list=PL848F2368C...