Gen Z asks senior developers about the future of programming

Tech/product themes discussed

• Shift from manual coding to “agentic”/AI-assisted workflows

  • Developers increasingly describe writing code in English via prompts/voice messages instead of typing commands character-by-character.
  • Even so, agent-generated changes still require validation—especially in large, high-stakes codebases, where architectural mistakes or incorrect generated code can be dangerous.
  • Teams still use line completion, and manual work remains common depending on complexity and risk.

• Higher-level abstraction trend (“English” as a programming layer)

  • Speakers connect historical abstraction jumps (e.g., HTML → frameworks like Angular → higher-level languages) to the idea that the next abstraction layer is natural language.
  • This may simplify development for many tasks, potentially increasing economic growth and job creation (with hope of more IT roles overall).

• Impact on junior developers: faster feedback and reduced time-to-prototype

  • Tools can reduce time from idea → working prototype (e.g., days/hours rather than weeks/months, including debugging and deployment).
  • This changes expectations: juniors can become productive sooner, narrowing the productivity gap with seniors in some areas—though it may also create more competition as access becomes easier.

---

Reviews, guides, tutorials, and “how-to” advice highlighted

• Educational path: university vs self-learning

  • University is defended mainly for:
    • Community and access to people/ambitious peers (not just content)
    • Fundamentals and depth (e.g., understanding low-level concepts that prevent treating tech as a “black box”)
  • Critique: curricula can become outdated because universities lag behind tool changes.

• How to build deep understanding in an “agent era”

  • Even with agents, fundamentals remain important:
    • Understanding LLMs/agents: why hallucinations happen, when to rely on an LM vs use web search, and limitations like knowledge cutoffs
  • Don’t treat AI tools as magic—understanding makes use more reliable and the work more interesting.

• Learning strategy under rapid tool change (“investment game”)

  • Use time like an investment:
    • Prefer evergreen fundamentals
    • Avoid “local hype” that becomes outdated quickly
  • Because content changes quickly, speakers recommend periodically trying things hands-on rather than relying only on blog posts or theory.

• Testing/QA with agents (practical approach)

  • Claims about current capability: general coding agents can use browser automation (e.g., Playwright) to:
    • open sites, click, fill forms, take screenshots
    • execute user-like test flows automatically
  • Suggested workflow:
    1. Build an agent-driven test plan
    2. Run the agent in the browser to validate scenarios
    3. Learn manual QA first, then replicate it using agent automation

• Practical “try it yourself” recommendation (repeated)

  • For claims like “agents can replace QA” or “describe a bug and AI fixes it,” speakers recommend:
    • spend ~30 minutes building a small experiment
    • implement an agent that checks a simple real scenario
    • validate feasibility given what’s available now

• How to aim for senior-level growth

  • Since “junior vs senior” blurs, the recommended path is to pursue senior-type tasks:
    • tackle complex, higher-level problem solving
    • teach yourself and build real systems (often via hobby projects)
  • Emphasis on end-to-end pipeline experience (frontend/backends + DB), plus agentic skills demonstrated in projects to show competence in interviews.

• Managing AI subscription costs

  • Suggestions include cheaper/open-source alternatives and services with generous educational plans.
  • Encouragement to ask for student access where possible.

---

AI/LLM concepts and risk framing

• Hallucinations and reliability

  • LLMs are described as predictable next-token systems.
  • Understanding helps you:
    • identify when output is trustworthy
    • know when to use web search or other grounding

• AI safety concerns are “practical”

  • Less emphasis on “AI will kill us” framing.
  • More focus on security issues, operational unknowns, and mitigation approaches.
  • Greater technical understanding correlates with fewer blanket fears and more concrete, actionable risk thinking.

---

Agentic tools & Open Source economics (problem + outlook)

• Problem: Agent-generated PRs/issues may flood maintainers, overwhelming review capacity.
• Observation: Some repositories shut down or limit contributions due to review burden and risk of vulnerabilities.
• Outlook: Speakers expect long-term adaptation:
  • More security/review challenges will appear (code review in an agentic universe, dependency management, secure execution).
  • Likely outcome: these challenges imply more demand for software engineers, not fewer.

---

Workforce/role evolution themes

• Manual front-end QA may change, but not entirely disappear

  • Agents can automate many test steps, but humans likely remain involved in planning, verifying, and ensuring coverage.

• Job market concern reframed

  • The short-term “junior hiring crisis” is treated as real (fewer opportunities since ~2022/2023 per discussion).
  • However, the barrier between junior and senior is expected to shift rather than vanish.
  • The learning curve is faster due to tooling and self-guided practice.

---

Advice on personal sustainability (“soft skills” and burnout)

• AI productivity can increase context switching and burnout risk

  • Agents blur work/life boundaries (e.g., quick prompts leading to evening/weekend work).

• Need for sustainable workflows

  • Manage energy/attention, stop when appropriate, and build habits around focus.
  • Experiment with what energizes/depletes you (e.g., sleep, gym/walks, meditation, stretching, limiting tech during low-energy times).

---

Main speakers / sources

• Dennis (senior developer)
• Andrew (senior/principal-type developer)
• Mentioned organizations/products: JetBrains, Jeff Brand Academy (example platforms), ChatGPT (with web search), Playwright, and various coding agents/LLMs (e.g., Claude/Gemini referenced via comparison)