Geoffrey Huntley 访谈：软件开发与软件工程的分野、开源之死与软件过剩时代

A couple of days ago, I sat down with Vivek Bharathi and dumped my brains. Here's the interview...

Below you'll find an AI transcription of everything we riffed about.

Key distinction: Software Development vs. Software Engineering:

Software development (typing code, prompting LLMs) is accelerating massively and becoming ubiquitous—anyone (e.g., a hairdresser using Cursor) can now be a "developer" due to abundant AI knowledge/tools.
Software engineering remains essential and is evolving: engineers now act like locomotive engineers — keeping the "train" on tracks by designing safe, reliable systems/automations rather than working "in" the business (manual coding).
Shift focus to designing loops, automations, safety mechanisms (e.g., sandboxing, credential management, security), risk engineering, and responsible AI utilization.

Implications for professionals:

If your identity is tied to being a traditional "software developer" (keyboard typing), it's a tough time—prompting for outcomes is the new norm.
If your employer bans AI tools, leave immediately: it's business suicide to ignore AI, while staying risks employability suicide as the market for manual coders shrinks rapidly.
Engineers should prioritize raw technical/cognitive skills → engineer away concerns (e.g., replace binary code reviews with risk-based approaches, feature flags, constrained blast radius, auto-migrations).

Traditional open-source libraries existed to ease hiring and sharing reusable code.
Now, with AI generation, there's little point: generating code avoids maintainer burnout, GitHub issue delays, abandoned projects, supply-chain attacks (e.g., npm takeovers), and Dependabot update toil.
Better to generate first-party code for faster evolution, full control, and no human "tool calls" (which disqualifies true AGI-like autonomy).
Exceptions: highly sensitive areas like PKI/SSL where generation isn't appropriate.

Software moves from scarcity (differentiated libraries, hard-to-replicate tech) to abundance (easy generation/reimplementation).
Many software products become hyper-commodity (like utilities: electricity, web hosting) — easily screenshot + reimplemented via AI (e.g., Claude).

True moats now lie in non-technical areas: contracts, relationships, handshakes, stakes, distribution, taste/judgment — the "hard things of business."
Unit economics of software have fundamentally changed → questions if software remains investable (VCs unsure about moats, fundraising challenges).

Future: hyper-personalized software; old models of building/scaling via scarcity are disrupted.

Stay relevant by running fast, staying curious, and adapting to the "brave new world."