Reflecting on Tim O'Reilly's "end of coding as we know it" prediction and what actually changed.
Change is slower than predicted
- A year after O'Reilly's essay: models got dramatically better, big tech friends code "completely differently now" — yet businesses operate in wait-and-see mode; most people are still employed, doing more work than before
1:01:46,1:02:08 - Companies over-rotated toward AI and are rebalancing — it's still trade-offs, still "move fast and break things"
1:02:40
What won't go away
- Business judgment, knowing what value to produce, and weighing engineering trade-offs: "how can you have an LLM weigh the trade-offs for you?"
1:03:06 - Host's addition: even a "simple" feature (a bug-report button) hides edge cases, state, user tiers, business domain — all context someone must hold and prioritize even when prompting instead of coding
1:03:29,1:03:47
Willingness to change as a survival trait
- Microsoft's voluntary buyouts were reportedly offered only to employees of a certain age with 10–15 years at the company — Navi speculates it targets people less willing to learn a completely new way of working; similar to the 2014 layoffs when Satya took over
1:04:32,1:04:56,1:05:17
No programmer extinction
- Every prior leap (cloud computing, higher-level languages) should have reduced programmer headcount — "it just hasn't happened"
1:05:43,1:06:11 - Counter-current to ever-higher abstraction: DHH/37signals left AWS (a few million dollars/year) for their own infrastructure and saved massively — at a certain scale, rolling your own stack pays
1:07:09,1:07:33 - Engineering adoption is largely culture, sales, and marketing (why MongoDB got big), and it snowballs — then corrects
1:07:59,1:08:18 - Warning: when AI subsidies run out, companies that embraced AI programming may cut back — cost will become "a really big issue"
1:08:43