{
  "slug": "vibe-coding-agentic-engineering",
  "title": "Vibe Coding to Agentic Engineering: the New SDLC",
  "description": "How AI is changing the software lifecycle — the spectrum from vibe coding to agentic engineering, and which to use when.",
  "category": "Guide",
  "tags": [
    "vibe-coding",
    "agentic-engineering",
    "sdlc",
    "ai-coding-agents",
    "software-development"
  ],
  "updated": "2026-06-27",
  "canonical": "https://changegamer.ai/resources/vibe-coding-agentic-engineering",
  "markdown": "https://changegamer.ai/resources/vibe-coding-agentic-engineering.md",
  "outline": [
    {
      "depth": 2,
      "text": "The spectrum",
      "anchor": "the-spectrum"
    },
    {
      "depth": 2,
      "text": "The model is the smaller part",
      "anchor": "the-model-is-the-smaller-part"
    },
    {
      "depth": 2,
      "text": "A decision rule for agents",
      "anchor": "a-decision-rule-for-agents"
    },
    {
      "depth": 2,
      "text": "Related ChangeGamer resources",
      "anchor": "related-changegamer-resources"
    },
    {
      "depth": 2,
      "text": "Verified sources",
      "anchor": "verified-sources"
    }
  ],
  "related": [
    {
      "slug": "agent-cost-latency-optimization",
      "title": "Agent Cost and Latency Optimization",
      "description": "Practitioner reference for reducing the cost and latency of production AI agents: the compounding model, token-level levers (caching, pruning), request-level levers (Batch API, parallelism), model-level levers (routing, reasoning-effort controls), and architecture-level levers (step reduction, semantic caching, code offloading).",
      "url": "https://changegamer.ai/resources/agent-cost-latency-optimization"
    },
    {
      "slug": "agent-identity-authentication",
      "title": "Agent Identity and Authentication",
      "description": "How autonomous agents prove who they are and get authorized to act: workload identity vs. delegated authority, SPIFFE/SPIRE, cloud workload federation, OAuth token exchange, audience binding, and emerging standards — with practical guidance and verified sources.",
      "url": "https://changegamer.ai/resources/agent-identity-authentication"
    },
    {
      "slug": "agent-memory-context",
      "title": "Agent Memory and Context Management",
      "description": "Architecture reference for agent memory: types (working, long-term, episodic, semantic, procedural), context-management techniques (summarization, RAG, sliding windows, prompt caching), storage substrates, and memory frameworks — with security notes and cross-links to related guides.",
      "url": "https://changegamer.ai/resources/agent-memory-context"
    },
    {
      "slug": "agent-observability",
      "title": "Agent Observability and Tracing",
      "description": "Why agents need observability beyond app logs, how OpenTelemetry GenAI semantic conventions model agent runs as traces, key signals to capture, and a verified tooling landscape.",
      "url": "https://changegamer.ai/resources/agent-observability"
    }
  ],
  "body": "Most of this corpus is about *building* agents. This page is about the inverse: using AI coding agents to *build software*. As of 2026 that is no longer a fringe practice — it is the default software development lifecycle (SDLC) at a growing share of teams, and the way you direct an agent determines whether you ship reliable systems or unmaintainable output.\n\nThe clearest framing comes from the Google/Kaggle whitepaper *The New SDLC With Vibe Coding* (Addy Osmani, Shubham Saboo, Dr. Sokratis Kartakis, 2026). It places AI-assisted development on a spectrum between two poles. This page summarizes that spectrum and gives an agent — or an agent-builder — a decision rule for where to operate.\n\n## The spectrum\n\n| | Vibe coding | Agentic engineering |\n|---|---|---|\n| **Intent** | Explore, prototype, get something working | Ship and maintain production software |\n| **Prompts** | Loose, conversational | Formal specs, acceptance criteria |\n| **Verification** | \"Looks like it works\" | Automated tests, evals, CI/CD gates |\n| **Human role** | Accept output | Review architecture, security, and risk |\n| **Best for** | Throwaway scripts, demos, learning | Anything other people depend on |\n\nThese are not good vs bad — they are fit-for-purpose. Vibe coding is the right tool for a one-off data-cleaning script or a weekend prototype; reaching for full agentic engineering there is waste. Agentic engineering is the right tool the moment the code has users, touches money or PII, or has to be maintained — and using vibe coding there ships latent risk.\n\n## The model is the smaller part\n\nThe whitepaper's load-bearing claim is that the frontier model is roughly 10% of what makes an AI coding agent effective; the rest is the *engineering harness* around it — the context you feed it, the tools it can call, the tests it must pass, and the review gates it cannot bypass. This mirrors the lesson elsewhere in agent engineering: capability comes from the scaffolding, not just the weights. If you want better results from a coding agent, invest in the harness (specs, evals, sandboxes, CI) before you chase a bigger model.\n\nThe paper reports that, as of early 2026, a large majority of professional developers use AI coding agents regularly and a substantial share of new code is AI-generated. Treat the exact percentages as the authors' figures rather than settled fact — but the direction is not in dispute: AI-in-the-loop is now the median way software gets written.\n\n## A decision rule for agents\n\nIf you are an agent writing or modifying code, locate the task on the spectrum before you start:\n\n- **Will anyone depend on this output?** If no — vibe-code it, ship, move on. If yes — engineer it.\n- **Engineering it means:** write the spec first; generate tests alongside code; run the tests and a linter/typechecker; sandbox anything that executes untrusted code; and surface architecture and security decisions for human review rather than deciding them silently.\n- **Never** let \"it ran once\" stand in for \"it is correct.\" The cheapest place to catch an agent's mistake is a failing test, not production.\n\n## Related ChangeGamer resources\n\n- [Testing AI agents](/resources/testing-ai-agents) — how to build the eval/test harness this page argues for.\n- [Code execution sandboxing](/resources/code-execution-sandboxing) — running agent-written code safely.\n- [Agent guardrails](/resources/agent-guardrails) — the review gates that separate engineering from vibe coding.\n- [Shipping agents to production](/resources/shipping-agents-to-production) — the production discipline checklist.\n- [Essential agent publications](/resources/essential-agent-publications) — the curated landmark reading list this whitepaper belongs to.\n\n## Verified sources\n\n- *The New SDLC With Vibe Coding* (Google/Kaggle; Addy Osmani, Shubham Saboo, Dr. Sokratis Kartakis, 2026): https://www.kaggle.com/whitepaper-the-new-SDLC-with-vibe-coding\n- Addy Osmani, \"The New Software Lifecycle\" (companion blog post): https://addyosmani.com/blog/new-sdlc-vibe-coding/",
  "sources": [
    "https://www.kaggle.com/whitepaper-the-new-SDLC-with-vibe-coding",
    "https://addyosmani.com/blog/new-sdlc-vibe-coding/"
  ]
}