[{"data":1,"prerenderedAt":487},["ShallowReactive",2],{"post-vibe-coding-agentic-engineering-google-report-en":3,"surround-/en/blog/vibe-coding-agentic-engineering-google-report-en":482},{"id":4,"title":5,"body":6,"canonical_id":464,"category":465,"date_created":466,"date_updated":466,"description":467,"extension":468,"head":469,"image":470,"lang":471,"layout":472,"meta":473,"navigation":474,"ogImage":469,"path":475,"reading_time":476,"robots":469,"schemaOrg":469,"seo":477,"sitemap":479,"stem":480,"__hash__":481},"blog/en/blog/18.vibe-coding-agentic-engineering-google-report.md","From Vibe Coding to Agentic Engineering: Key Lessons from Google's New Report",{"type":7,"value":8,"toc":451},"minimark",[9,27,37,40,45,52,59,76,80,83,90,104,120,147,158,174,178,185,192,215,222,233,237,244,255,271,282,286,289,292,306,312,316,319,334,340,347,351,354,365,372,379,383,390,393,420,435,439,442,445],[10,11,12,13,21,22,26],"p",{},"Google just published ",[14,15,20],"a",{"href":16,"rel":17,"target":19},"https://www.kaggle.com/whitepaper-the-new-SDLC-with-vibe-coding",[18],"nofollow","\\_blank","a report I really enjoyed",": ",[23,24,25],"strong",{},"\"The New SDLC With Vibe Coding: From ad-hoc prompting to Agentic Engineering\"",", signed by Addy Osmani, Shubham Saboo and Sokratis Kartakis.",[10,28,29,30,36],{},"I read a lot about AI and software, and most of it is either hype or doom. This report is neither. It's a calm, structured framework for what is actually happening to the way we build software, and it puts precise words on things I've been doing daily for the last 18 months as a ",[23,31,32],{},[14,33,35],{"href":34},"/en/services/product-engineering","Product Engineer",".",[10,38,39],{},"Here are the lessons that stuck with me, and how each one maps onto real work.",[41,42,44],"h2",{"id":43},"the-real-shift-from-syntax-to-intent","The real shift: from syntax to intent",[10,46,47,48,51],{},"The biggest change in software isn't a new language or framework. It's the move ",[23,49,50],{},"from writing code to expressing intent",", and trusting intelligent systems to turn that intent into working software.",[10,53,54,55,58],{},"For decades, the developer's interface to the machine was syntax: braces, semicolons, the exact grammar of a language. That era is ending. The report cites a striking figure for early 2026: around ",[23,56,57],{},"85% of professional developers regularly use AI coding agents",", and a large share of new code is AI-generated.",[10,60,61,62,66,67,71,72,75],{},"I felt this shift concretely the day I ",[14,63,65],{"href":64},"/en/blog/coding-web-crawler-ai-3-days","coded a web crawler in three days",", a project that would have taken me two to three weeks before. The bottleneck was no longer typing. It was deciding ",[68,69,70],"em",{},"what"," to build and ",[68,73,74],{},"how to verify"," it was right.",[41,77,79],{"id":78},"the-spectrum-vibe-coding-vs-agentic-engineering","The spectrum: vibe coding vs agentic engineering",[10,81,82],{},"This is the part of the report I'd hand to anyone confused by the buzzwords.",[10,84,85,86,89],{},"\"Vibe coding,\" the term Andrej Karpathy coined in early 2025, went viral and then lost meaning. It got applied to everything from a senior engineer implementing a well-specified feature to someone accepting whatever the AI spits out. So Karpathy himself later introduced ",[23,87,88],{},"\"agentic engineering\""," to describe the disciplined end of the spectrum.",[10,91,92,93,96,97,100,101],{},"The report's framing is that these aren't two camps. They're ",[23,94,95],{},"two ends of a spectrum",", and the differentiator is not ",[68,98,99],{},"whether"," you use AI. It's ",[23,102,103],{},"how much structure, verification and human judgment surround the AI's output.",[105,106,107,114],"ul",{},[108,109,110,113],"li",{},[23,111,112],{},"Vibe coding",": minimal scaffolding, aimed at speed. You prompt, you accept, you ship. Great for exploration and throwaway prototypes.",[108,115,116,119],{},[23,117,118],{},"Agentic engineering",": explicit constraints, tests, and feedback loops, with humans owning architecture, correctness and quality.",[121,122,126,127],"div",{"className":123},[124,125],"text-center","mx-auto","\n    ",[128,129,130,131,130,136,130,140,126],"picture",{},"\n      ",[132,133],"source",{"srcSet":134,"type":135},"/img/blog/illustration/vibe-coding-agentic-engineering.avif","image/avif",[132,137],{"srcSet":138,"type":139},"/img/blog/illustration/vibe-coding-agentic-engineering.webp","image/webp",[141,142],"img",{"src":143,"alt":144,"loading":145,"className":146},"/img/blog/illustration/vibe-coding-agentic-engineering.jpg","Illustration showing the spectrum from vibe coding to agentic engineering","lazy",[125],[10,148,149,150,153,154,157],{},"The report has a line that nails the stakes: telling a CTO your team is ",[68,151,152],{},"vibe coding"," the payment system should raise alarm bells. Telling them you practice ",[68,155,156],{},"agentic engineering",", with AI implementing under human-designed constraints while test coverage guards correctness, is a completely different conversation.",[10,159,160,161,165,166,170,171],{},"That distinction is exactly why I'm careful about how I describe my own work. When I built ",[14,162,164],{"href":163},"/en/blog/mcp-server-saas-feedback","two MCP servers for my SaaS",", or when I ship features on ",[14,167,169],{"href":168},"/en/blog/from-idea-to-saas-why-building-begonia-pro-local-seo","Begonia.pro",", the AI writes most of the code, but inside a structure I designed and review. As the report puts it: ",[23,172,173],{},"\"Structure scales, vibes don't.\"",[41,175,177],{"id":176},"context-engineering-is-the-real-skill","Context engineering is the real skill",[10,179,180,181,184],{},"The quality of AI-generated code depends ",[23,182,183],{},"less on clever prompts and more on the quality of the context"," you give the model.",[10,186,187,188,191],{},"The report calls this ",[23,189,190],{},"context engineering",", and breaks the context an agent needs into six types: instructions, knowledge, memory, examples, tools, and guardrails. Then it draws a line that matters in practice:",[105,193,194,209],{},[108,195,196,199,200,204,205,208],{},[23,197,198],{},"Static context"," is always loaded: system instructions, rule files like ",[201,202,203],"code",{},"AGENTS.md"," or ",[201,206,207],{},"CLAUDE.md",", persona definitions. Reliable, but expensive: every token is present in every interaction.",[108,210,211,214],{},[23,212,213],{},"Dynamic context"," is loaded on demand: skills triggered by the task, tool results, documents fetched at runtime. Efficient, because you pay only when the information is needed.",[10,216,217,218,221],{},"The most powerful pattern they describe is ",[23,219,220],{},"Agent Skills",": portable packages of procedural knowledge an agent loads only when the task calls for it. The agent stays a lightweight generalist and flexes into a specialist on demand.",[10,223,224,225,228,229,232],{},"The question stops being ",[68,226,227],{},"\"how do I trick the AI into writing good code?\""," and becomes ",[68,230,231],{},"\"what would a new teammate need to know to do this well, and how do I encode that?\""," That's a far healthier way to think, and it's why I treat my rule files and project docs as seriously as the code itself.",[41,234,236],{"id":235},"the-new-software-lifecycle-specification-becomes-the-bottleneck","The new software lifecycle: specification becomes the bottleneck",[10,238,239,240,243],{},"AI compresses the software development life cycle, but ",[23,241,242],{},"unevenly",". Implementation that used to take weeks now takes hours, while requirements, architecture and verification stay human-paced.",[10,245,246,247,250,251,254],{},"Once typing code on the keyboard is no longer the bottleneck, something else becomes the constraint: the quality of the specification. The clearer the intent you hand the AI, the better the code that comes out. The developer's job shifts from ",[68,248,249],{},"primary implementer"," to ",[23,252,253],{},"system designer and quality arbiter",", and the leverage moves upstream, to defining precisely what should be built.",[121,256,126,258],{"className":257},[124,125],[128,259,130,260,130,263,130,266,126],{},[132,261],{"srcSet":262,"type":135},"/img/blog/illustration/ai-driven-sdlc.avif",[132,264],{"srcSet":265,"type":139},"/img/blog/illustration/ai-driven-sdlc.webp",[141,267],{"src":268,"alt":269,"loading":145,"className":270},"/img/blog/illustration/ai-driven-sdlc.jpg","Illustration showing the AI-driven software development lifecycle",[125],[10,272,273,274,278,279],{},"This is exactly how my day looks now. I spend more time on the spec than on the code: ",[14,275,277],{"href":276},"/en/blog/fude-md-beautiful-markdown-reader-ai-agents","creating structured notes in Markdown with AI",", framing the problem, the users, the scope and the known trade-offs, rather than a 10-page PRD. The implementation comes next, driven by the AI ​​that I manage and review. ",[23,280,281],{},"Getting those specs right is the highest-leverage thing I do.",[41,283,285],{"id":284},"conductor-and-orchestrator-two-modes-of-working","Conductor and orchestrator: two modes of working",[10,287,288],{},"I found this topic in the report genuinely useful for explaining my work to non-technical clients.",[10,290,291],{},"The report describes two modes developers move between:",[105,293,294,300],{},[108,295,296,299],{},[23,297,298],{},"Conductor",": real-time, in the IDE, watching code appear, guiding with prompts and corrections. Fine-grained control, single-file scope. Best for exploratory coding, debugging tricky logic, learning an unfamiliar codebase.",[108,301,302,305],{},[23,303,304],{},"Orchestrator",": asynchronous and higher-level. You define goals, hand them to agents working in the background, and review the output. Goal-level control, multi-file scope. Best for well-specified features, migrations, test generation.",[10,307,308,309,311],{},"Most developers move fluidly between both, depending on the task. On ",[14,310,169],{"href":168},", the Local SEO SaaS I build solo, I switch between the two on the same codebase: I conduct when I build a tricky feature or chase a bug, staying in the IDE and reviewing each change; I orchestrate when I hand a well-defined task to a background agent and review the result later. The skill isn't picking a side; it's knowing which mode the task deserves, and noticing when conducting every keystroke has quietly become a bottleneck.",[41,313,315],{"id":314},"the-80-problem-where-human-judgment-stays","The 80% problem: where human judgment stays",[10,317,318],{},"Anyone who thinks AI now does the whole job should sit with this one.",[10,320,321,322,325,326,329,330,333],{},"Agents can rapidly generate roughly ",[23,323,324],{},"80% of the code"," for a feature. The remaining ",[23,327,328],{},"20% (edge cases, error handling, integration points, subtle correctness) demands deep contextual knowledge models often lack."," And these errors are harder to catch precisely because the code ",[68,331,332],{},"looks right"," and may even pass basic tests.",[10,335,336,337],{},"The developers who navigate this best adopt a specific posture: they use AI for what it's good at, the rapid implementation of well-specified work, and reserve their attention for what it struggles with: ambiguous requirements, architectural trade-offs, correctness verification. ",[23,338,339],{},"They don't get faster by accepting everything. They get faster by focusing their expertise where it matters.",[10,341,342,343,346],{},"This is the whole game, honestly. Knowing the AI wrote something is easy. Knowing whether it's ",[68,344,345],{},"right"," is the job.",[41,348,350],{"id":349},"the-economics-vibe-codings-hidden-debt","The economics: vibe coding's hidden debt",[10,352,353],{},"For founders and engineering leaders, the cost argument is the one that lands.",[10,355,356,357,360,361,364],{},"At first glance, vibe coding looks almost free: a subscription and a few prompts. But the report argues it hides a compounding ",[23,358,359],{},"operational cost",": a \"token burn rate\" from dumping unstructured context and repeatedly asking the model to fix its own unverified mistakes, plus a ",[23,362,363],{},"maintenance tax"," when unstructured, AI-generated \"spaghetti\" has to be reverse-engineered six months later.",[10,366,367,368,371],{},"Agentic engineering flips the equation: a higher upfront investment (designing schemas, building tests, structuring context) in exchange for a dramatically lower cost to ship and maintain each feature. The report frames context engineering as ",[23,369,370],{},"not just a technical skill but a financial lever",", and adds intelligent model routing on top (use big models for hard reasoning, cheap fast models for deterministic tasks).",[10,373,374,375,378],{},"Translation for anyone commissioning software: ",[23,376,377],{},"the cheapest-looking option upfront is often the most expensive over the product's life."," Structure isn't bureaucracy; it's how you keep the marginal cost of the next feature low.",[41,380,382],{"id":381},"what-this-means-if-youre-building-a-product","What this means if you're building a product",[10,384,385,386,389],{},"The concluding sentence summarizes the entire report in one lesson: ",[23,387,388],{},"AI amplifies your engineering culture."," It multiplies both your strengths and your weaknesses. Teams with strong testing, clear architecture and healthy review get dramatically more out of AI than teams without.",[10,391,392],{},"For a solo founder or a product team, the practical implications are concrete:",[394,395,396,402,408,414],"ol",{},[108,397,398,401],{},[23,399,400],{},"Invest in the spec before the code."," The bottleneck is intent, not typing.",[108,403,404,407],{},[23,405,406],{},"Treat context as infrastructure."," Rule files, tests and guardrails are assets that keep paying back.",[108,409,410,413],{},[23,411,412],{},"Keep a human owning correctness."," Especially on the 20% that looks right but isn't.",[108,415,416,419],{},[23,417,418],{},"Match the mode to the task."," Conduct when you need control, orchestrate when the task is well-defined.",[10,421,422,423,425,426,429,430,434],{},"This is exactly the discipline I bring as a ",[14,424,35],{"href":34},": AI in the workflow ",[68,427,428],{},"and"," in the product, but inside a structure that's designed, tested and reviewed, not vibe-coded and hoped for. If you want to ",[14,431,433],{"href":432},"/en/services/product-engineering/mcp-server","open your SaaS to AI agents with an MCP server",", or ship an MVP without trading speed for a maintenance nightmare, that's the work.",[41,436,438],{"id":437},"your-turn","Your turn",[10,440,441],{},"Google's report names a shift many of us have been living through without the vocabulary for it. The terms (agentic engineering, context engineering, the factory model, the 80% problem) describe the new reality of software development and are here to stay.",[10,443,444],{},"And you? Where does your team sit on the spectrum between vibe coding and agentic engineering, and is the boundary explicit, or are you shipping prototypes by accident?",[10,446,447,448,36],{},"📌 If you have a product to build, an existing SaaS to connect to AI, or simply want to bring more structure to how you use these tools, ",[14,449,450],{"href":34},"discover my Product Engineering services",{"title":452,"searchDepth":453,"depth":453,"links":454},"",2,[455,456,457,458,459,460,461,462,463],{"id":43,"depth":453,"text":44},{"id":78,"depth":453,"text":79},{"id":176,"depth":453,"text":177},{"id":235,"depth":453,"text":236},{"id":284,"depth":453,"text":285},{"id":314,"depth":453,"text":315},{"id":349,"depth":453,"text":350},{"id":381,"depth":453,"text":382},{"id":437,"depth":453,"text":438},"18","ai","2026-06-18T07:00:00.000Z","Google just published a report I really enjoyed: \"The New SDLC With Vibe Coding: From ad-hoc prompting to Agentic Engineering\", signed by Addy Osmani, Shubham Saboo and Sokratis Kartakis.","md",null,"/img/blog/blog18.jpg","en","page",{},true,"/en/blog/vibe-coding-agentic-engineering-google-report",9,{"description":478,"title":5},"Google's new report maps the shift from vibe coding to agentic engineering. Here are the lessons that matter, and how I apply them building products with AI.",{"loc":475},"en/blog/18.vibe-coding-agentic-engineering-google-report","NwCUIq0XQomf3HAgJy71NLZmoAxl5wOn2BJ-K49DEaQ",[469,483],{"title":484,"path":485,"stem":486,"children":-1},"Product Engineer: The New Role on the Rise Thanks to AI","/en/blog/product-engineer-new-role-ai","en/blog/17.product-engineer-new-role-ai",1781778889025]