In 1987, clustered regularly interspaced palindromic repeats or CRISPR was accidentally discovered in the sequences of DNA in E coli bacteria.
It was not until 2012, that Doudna and others directed CRISPR into a precise gene editing tool.
In the first chapter of When We Cease to Understand the World, they tell the story of Carl Wilhelm Scheele, a chemist who invented cyanide by combining sulphuric acid and Prussian Blue. Again, it is not the discovery of cyanide that was notable but the optimization of it into a dangerous gas used infamously on the trenches of World War I.
I've been thinking a lot recently about how discovery is distinctly accidental and comes years before the engineering and packaging that eventually makes it impactful.
Optimization rarely appears to lead to discovery.
I work at a company where we are building a search engine. We're sitting on the backbones of years of crawling, indexing research efforts. But time and time again, all the actual value we generate as a company comes from variations of packaging: product interfaces, API design, "Exa for code search". The abstraction of the search engine needs to be legible for people to use, the models do not in fact sell themselves.