Add "we can't code good (so take LLM hype with a grain of salt)"

diff --git a/programming/we_cant_code_good_so_take_llm_hype_with_a_grain_of_salt.md b/programming/we_cant_code_good_so_take_llm_hype_with_a_grain_of_salt.md
new file mode 100644 (file)
index 0000000..675f8a2
--- /dev/null
+++ b/programming/we_cant_code_good_so_take_llm_hype_with_a_grain_of_salt.md
@@ -0,0 +1,32 @@
+# We can't code good (so take LLM hype with a grain of salt)
+
+Software engineering is a young field compared to most other fields.
+
+All studies about software engineering projects I have seen say that most projects are late, over budget, and underdeliver.
+
+(A great article did some statistics to estimate the percentage of software engineers who have never seen worked on a successful project.
+The result was discouraging.
+I have never found this article again, please contact me if you know it.)
+
+Consider CRUD applications, which most software engineers consider to be the lowest kind of work.
+A massive amount of CRUD applications have been developed, and [I believe we are not effective at churning them out](crud_is_an_important_unsolved_problem.md).
+
+[What we know we don't know: empirical software engineering](https://www.hillelwayne.com/talks/ese/ddd/) is an interesting talk by Hillel Wayne that elaborates on this topic.
+
+The talk claims that researchers who try to prove that any engineering practice has a significant effect on delivery mostly fail.
+Apparently, code review seems to be one of the few practices that has been proven to be effective but its effectivity is small compared to sleeping well, not being stressed, or working the right amount of hours.
+
+This includes practices that most of us believe have significant effects, such as using dynamic or static typing, abbreviating or not identifiers, and many others.
+
+We all believe that some practices are effective, yet this does not seem to be measurable.
+
+I think experience, good team work, smarts, and motivation help.
+(And I believe there are certainly things that *hinder* software engineering!)
+We have been able to deliver good works of engineering, and some things seem to work, but we can't code good yet.
+
+So *any* claim of *any* practice being effective should be taken with a dose of healthy skepticism.
+I do not even know how to measure software engineering effectivity; I can use my experience to compare and try to infer conclusions, but I know my conclusions can be wrong.
+
+I am sure that at some point our discipline will be as mature as other disciplines, but think how much effort it took other disciplines to get there.
+But until then, we should be humble and not make bold claims, and we should question the incentives of those who make bold claims.
+Perhaps some of those bold claims turn out to be true, but the longer that you have been in this field, the more bold claims you will have seen fail.