I no longer care to work on any Ruby gems so I've archived all my repositories
A longtime Ruby developer has archived all their open-source gems, arguing that large language models are strip‑mining human-written code and turning programming into low-quality, AI-generated “slop.” Commenters are sharply divided: some share a sense of grief and demotivation as hand-crafted coding and open-source community work feel devalued, while others see LLMs as powerful tools that democratize software creation and free engineers from repetitive tasks. Underneath the language and Ruby-specific drama are broader worries about job security, code quality, regulatory capture, and whether future software will remain understandable and meaningfully human-made.
Emotional response and loss of joy
- Many long‑time programmers, especially those who valued code as a craft, describe grief, demotivation, and a sense that “the fun is gone.”
- Ruby is seen as emblematic of “joyful coding,” so LLM‑first messaging around Ruby/Rails feels like a betrayal of that culture.
- Several compare the shift to moving from artisan woodworking or tailoring to monitoring factory robots; you can still practice the craft, but mostly as a non‑economic hobby.
AI as tool vs replacement
- Some feel liberated: LLMs remove drudgery (CRUD, boilerplate, debugging glue), let individuals build ambitious projects, and enable non‑experts to create software.
- Others see this as hollowing out the interesting parts of the job, turning engineers into “agent managers” or product managers, which they never wanted to be.
- There’s disagreement on whether using agents is itself a legitimate, enjoyable craft or inherently a degradation of programming.
Open source, licensing, and “strip‑mining”
- Strong sentiment that open source has been “strip‑mined” for LLM training: unpaid labor scraped, mixed with proprietary code, and sold back as API access.
- Some argue this was always true of corporate use of OSS; others see LLM training as qualitatively worse.
- Ideas floated: stronger copyleft, new GPL versions “AI‑proof,” and government‑funded public software; skepticism that any of this will restrain large vendors.
Quality, maintainability, and “vibe coding”
- Widespread worry that AI encourages shallow, non‑deterministic “vibe coded” systems: code that works today but is fragile, incoherent, and hard to reason about.
- Concerns that maintainers will drown in low‑effort AI PRs and bug reports, pushing more of them to archive or abandon projects.
- Counterpoint: with good discipline, LLMs can improve refactoring, testing, and integration; some claim better coverage and tooling on their personal projects than at work.
Economics and labor market
- Analogies: cheap robots vs master carpenters; factoryization of software; factory workers and miners replaced by machines.
- Some assert order‑of‑magnitude cost reductions for many tasks; others say true costs (R&D, infra, externalities) are unknown and likely high.
- Fear that SWE salaries and headcount will fall as agent workflows mature; calls for unionization, but skepticism it would change much.
Regulation, power, and open models
- Discussion of regulatory capture scenarios: national bans on specific models, pressure on hardware supply, and use of security fears to justify centralization.
- Hope that open‑weight, self‑hosted models will limit “rug pulls”; countered by concerns about regulation and compute chokepoints.
Impact on software and ecosystems (so far)
- Some see little real renaissance: core tools (OSes, major apps, CAD, creative suites) are not obviously transformed, and many new AI‑built apps feel low‑quality.
- Others point to quieter changes: rapid reverse‑engineering, bespoke internal tools, ports, personal automation, and small niche projects that wouldn’t have existed before.
- Expectation that as human maintainers exit, ecosystems may suffer in stability and innovation, even while AI can churn out more code than ever.