No Silver Bullet at Forty

Coding models removed the typing. Brooks said that would not be enough.

In 1986, Fred Brooks presented a paper at the IFIP Congress with a title borrowed from werewolf folklore: “No Silver Bullet: Essence and Accident in Software Engineering”. Software projects, he observed, have a way of turning into monsters, running late, over budget and riddled with flaws, and everyone wants the one weapon that kills the monster outright. His verdict was that no such weapon was coming. Not within a decade, and not from any single technology or management trick.

The essay reached a wider audience in 1987 through IEEE Computer, and it has been reprinted and argued over ever since. It turns forty this year, which makes the timing awkward in the best way. We now have tools that write working code from a paragraph of English, and the question Brooks asked about every previous candidate is back on the table: is this the one?

My position is that coding models are the closest thing to a silver bullet the field has ever produced, and that this is precisely why Brooks was right. They remove the part of the work he called accidental more completely than anything before them. What remains is the part he said would remain, and it is as hard as he claimed. There is a serious counterargument, though, and it deserves a hearing…