# vibe coding: from prompts to parts — and why specs are taking over — Skills & Case Studies vibe coding is being pulled in three new directions this week: the same instinct is now designing physical parts (vibe manufacturing), the code-writing habit is shifting to spec-driven development, and formal verification just compressed months of expert work into half a day. Three original skills and three reported stories to make sense of it. ## Skills - Spec-Driven Agent Workflow: Write the spec first, then let the agent write the code. (https://aiservicesmarket.lol/skills/spec-driven-agent-workflow/) - AI Code Security Review: Review AI-generated code against the OWASP Top 10 before it ships. (https://aiservicesmarket.lol/skills/ai-code-security-review/) - Vibe Manufacturing Blueprint: From a one-line idea to a physical part: prompt, design, order, verify. (https://aiservicesmarket.lol/skills/vibe-manufacturing-blueprint/) ## Case studies - Ten Minutes, One Steel Part: Vibe coding has a hardware sibling: ‘vibe manufacturing,’ where a chatbot designs a part and the file goes straight to a manufacturer or a home 3D printer. The week’s example was Jeff Deehan, an East Lansing real-estate developer who had a chatbot design a steel part for GPT-6 Astra voice mode in about five minutes, downloaded the DXF, and ordered it from SendCutSend for $156 — it arrived ten days later. - Why Specs Are Replacing Vibes: The counter-argument to pure vibe coding got its sharpest form this week: specs first, code second. The numbers behind the shift are stark — Veracode’s 2025 data put OWASP Top 10 flaws in about 45% of AI-generated code, and a METR study found developers with AI tools took 19% longer, not less. The tooling answer is spec-driven: AWS Kiro (July 2025) writes requirements, design, and tasks first; GitHub’s Spec Kit (September 2025) formalizes specify/plan/tasks/implement. - Half a Day vs Five Months: The most dramatic data point for AI-assisted rigor: Stanford researcher Chuyue Sun used AI to formally verify part of Signal’s encryption code in under half a day for less than $500 — work that took three experts five months by hand. Days later, &Truth emerged from stealth with $70 million in seed funding to build a universal verification machine on the Lean proof system. ## Frequently asked questions ### What is vibe coding? Vibe coding means describing software in natural language and letting an AI agent write the code from that feeling. The term was coined by Karpathy in February 2025. If you want the discipline that is replacing the pure-vibe approach, see Spec-Driven Agent Workflow. ### Is vibe coding dying? A November 2026 essay argues the industry is moving from open-ended vibe coding to spec-driven development — write requirements, design, and tasks first, then let the agent implement. The quality data backs the concern: Veracode's 2025 report found roughly 45% of AI-generated code had OWASP Top 10 vulnerabilities. Run the new pattern with Spec-Driven Agent Workflow. ### What is vibe manufacturing? The same describe-it-and-let-AI-build-it instinct applied to hardware: a chatbot designs a part, you export a manufacturable file like DXF, and a fabricator or home 3D printer makes it real. The documented case: five minutes of design, a $156 SendCutSend order, delivery in ten days. The full pipeline is in Vibe Manufacturing Blueprint. ### How do I make AI-written code secure? Review it like it is guilty until proven innocent: trace every user-controlled value to its sink, check XSS first (an 86% miss rate per Veracode 2025), verify authorization on every new endpoint, and keep secrets out of logs. The complete checklist is AI Code Security Review. ### What is spec-driven development, concretely? Four stages: specify requirements, plan the design, break the work into tasks, implement task by task — the pattern behind AWS Kiro's requirements/design/tasks files and GitHub Spec Kit's specify/plan/tasks/implement. A ready-to-run worksheet is in Spec-Driven Agent Workflow. ### Why does AI coding need specs at all? Because plausibility is not correctness: beyond the 45% vulnerability figure, METR research found developers using AI tools took 19% longer while believing they were 20% faster. A written spec gives both you and the agent something checkable — see Spec-Driven Agent Workflow and verify with AI Code Security Review. ### Can AI really verify encryption code? In the reported case, yes for a defined slice: Chuyue Sun used AI to complete part of a Signal encryption-code verification in under half a day and under $500, versus five months of manual work by three experts. New companies like &Truth ($70M seed) are productizing exactly this. For everyday code, start with AI Code Security Review. ### Is vibe manufacturing practical for real products? For simple parts, the documented economics look real: a steel part for $156 in ten days. The reported limits are complexity and expensive trial-and-error — complex assemblies still need a human engineer, and you should budget for remakes. Plan for that with Vibe Manufacturing Blueprint. ### What tools support the spec-driven shift? AWS Kiro (July 2025) builds the requirements/design/tasks flow into the IDE, and GitHub Spec Kit (September 2025) adds specify/plan/tasks/implement stages. Both implement the same idea: write the spec first. A tool-agnostic version is Spec-Driven Agent Workflow. ### Where do I start if I am new to all of this? Start with the workflow, not the tools: Spec-Driven Agent Workflow for writing code from specs, AI Code Security Review before anything ships, and Vibe Manufacturing Blueprint if you want to hold your output in your hand. Part of the trending collections at https://aiservicesmarket.lol/trending/