Tell HN: I'm 60 years old. Claude Code has re-ignited a passion
Large language model coding agents like Claude Code are reigniting enthusiasm for software development, especially among older engineers and hobbyists who are suddenly able to ship long‑deferred projects or build full apps solo. Many describe a shift from hand‑coding to specifying architecture, constraints and tests while AI handles boilerplate, increasing throughput but also raising questions about code quality, craftsmanship and how to train new developers. Alongside optimism about democratizing software creation, there is deep unease about job security, dependency on a few AI providers, the erosion of hard‑won expertise, and the risk that organizations will prioritize short‑term velocity over maintainable systems.
Renewed passion and accessibility
- Many older developers (50s–70s) say Claude Code/LLMs have reignited their desire to build things, especially long‑deferred personal projects.
- People with health issues, burnout, ADHD, or reduced attention span describe AI as an “accessibility tool” that removes frustrating toil and lets them keep working.
- Several non‑programmers or casual scripters report building full apps for the first time, often for very personal, niche workflows.
How people are using agentic coding
- Common use cases: personal productivity apps (habits, health, inventory, media tracking), small SaaS tools, data pipelines, infra automation, trading/backtesting tools, educational tools, and game/toy projects.
- Typical workflow: human writes specs and breaks work into phases; agent generates code; human iterates, refactors, and reviews, sometimes with multiple models cross‑checking.
- Some use agents to glue together existing scripts/notebooks into cohesive apps, or to port old code/binaries into new stacks.
Shift in what “programming” means
- Many argue coding is becoming “LLM wrangling”: designing systems, specs, and architectures, then steering and verifying agents.
- Experienced devs say their value now lies more in judgment, domain knowledge, and architecture than in typing code.
- Others feel this devalues decades spent mastering languages, tooling, and low‑level debugging.
Fulfillment, craftsmanship, and learning
- Some find agentic coding exhilarating because it collapses idea‑to‑MVP time and removes boring repetition.
- Others feel hollow or “like cheating”: they miss flow, puzzle‑solving, and the pride of having written the code themselves.
- Concerns: weaker deep understanding, harder to maintain “ownership” of code, and difficulty cultivating craftsmanship when AI does the implementation.
Quality, reliability, and testing
- Reports range from “production‑grade daily” to “great for prototypes but breaks on complex refactors.”
- Frequent themes: need for strong tests, guardrails, and human oversight; agents can hallucinate APIs, over‑refactor, or introduce subtle bugs.
- Some compare LLMs to junior devs: fast, but require review, constraints, and good prompts to avoid slop.
Careers, democratization, and risk
- Optimists: AI democratizes software creation, empowers solo founders, and massively amplifies strong engineers.
- Pessimists: fear displacement of juniors, commoditization of coding, and concentration of power in a few AI vendors.
- Ongoing debate over IP/licensing of generated code, ethical training data, and whether this is a sustainable “golden age” or hype.
Meta and skepticism
- A visible minority suspects astroturfing, noting vague project descriptions and highly enthusiastic tone.
- Others counter with detailed project lists and argue that even if hype exists, the practical gains are real.