“Vibe coding is the new
product management.”
Training and tuning models is the new coding.
Naval Ravikant
@naval · 3M+ followers
Naval Ravikant
Angel Investor, Philosopher
"Vibe coding is the new product management."
Andrej Karpathy
Ex-Tesla AI, OpenAI
"Agentic engineering is the grown-up successor."
Tomer Cohen
CPO, LinkedIn
"We want your work, not your job history."
Andrew Ng
AI Pioneer, DeepLearning.AI
"The PM:Eng ratio is flipping to 2:1."
The Tweet That Reframed Product Management
On February 3, 2026, Naval Ravikant — angel investor, philosopher-CEO, and one of tech's most influential voices — posted 21 words to his 3 million+ followers on X. Within hours, it became the most debated take in the product management world.
The tweet landed exactly one year after Andrej Karpathy — former Tesla AI director and OpenAI founding member — coined the term “vibe coding” on February 3, 2025. What Karpathy described as a playful new way to interact with AI coding assistants — “fully give in to the vibes, embrace exponentials, and forget that the code even exists” — had, in 12 months, evolved into something far more consequential.
Naval's reframing wasn't just about coding. It was about power. If anyone with product judgment can now instruct AI to build software, the traditional PM role — the person who writes requirements for engineers to implement — is no longer the bottleneck. The person with the vision can now build it directly. And that changes everything.
The next day, Aakash Gupta — one of the most-followed PM voices on X — posted a thread unpacking the implications:
Timeline: From Vibe Coding to PM Infrastructure
The journey from a “throwaway tweet” to an industry-redefining concept took exactly one year. Here is how it happened.
Feb 2025
Karpathy coins "vibe coding"
Mid 2025
LinkedIn launches APB program
Late 2025
Collins Dict. Word of Year
Jan 2026
McKinsey: 11% AI ROI
Feb 3, 2026
Naval's tweet
Feb 4, 2026
Karpathy: "agentic eng"
Karpathy coins "vibe coding"
Andrej Karpathy tweets: "There's a new kind of coding I call vibe coding, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists." The tweet gets 4M+ views.
LinkedIn launches APB program
LinkedIn's CPO Tomer Cohen replaces the traditional APM program with the Associate Product Builder (APB) program, fusing PM, design, and engineering into a single role.
Collins Dictionary: "vibe coding" is Word of the Year
Collins Dictionary names "vibe coding" its 2025 Word of the Year. Stack Overflow survey finds 84% of developers vibe code or plan to start; 47% do it daily.
AI PM tools go mainstream
ChatPRD, Amplitude AI, BuildBetter, and other AI-native PM tools move from experimental to standard infrastructure across product organizations.
McKinsey reports AI adoption gap
43% of companies report productivity gains from AI, but only 11% have realized measurable ROI at scale — a "product management gap" where organizations cannot translate AI capability into user value.
Naval tweets: "Vibe coding is the new PM"
Naval Ravikant posts to 3M+ followers: "Vibe coding is the new product management. Training and tuning models is the new coding." The tweet triggers an industry-wide conversation.
Karpathy's one-year retrospective
Karpathy reflects on vibe coding's anniversary: "This was a shower of thoughts throwaway tweet that somehow minted a fitting name at the right moment." Proposes "agentic engineering" as the mature successor.
Aakash Gupta frames the PM angle
"Naval just told 3M+ people that PMs who can vibe code are the new power players in tech... The gap between someone who can vibe code and someone who can vibe code with product judgment is the gap between a demo and a business."
The Data Behind the Shift
Naval's tweet wasn't hot air — it was backed by a convergence of data points that all pointed in the same direction.
of developers vibe code or plan to
Stack Overflow 2025 Developer Survey
vibe code every single day
Stack Overflow 2025 Developer Survey
of all code is now AI-generated
Stack Overflow 2025 Developer Survey
of product teams report more accurate decisions with AI analytics
BuildBetter.ai Research, 2025
of companies report productivity gains from AI
McKinsey Global AI Report, 2024
have realized measurable ROI at scale
McKinsey Global AI Report, 2024
100%
Companies investing in AI
43%
Report productivity gains
11%
Achieve measurable ROI
32-Point Gap = PM Problem
The gap between AI capability and AI ROI is not a technology problem. It is a product management problem. Organizations need product thinking to translate AI into user value.
The AI-ROI Gap Is a PM Problem
The 32-point gap between “companies seeing AI productivity gains” (43%) and “companies achieving measurable ROI” (11%) is not a technology problem — it is a product management problem. Organizations can deploy AI, but they cannot translate that capability into user value at scale without product thinking. This is precisely why PM skills are becoming the critical infrastructure layer for enterprise AI adoption.
What Actually Changed for PMs
Document Writer
Traditional PM (2020-2024)
- •Write PRDs & specs
- •Manage Jira backlog
- •Run stakeholder meetings
- •Request data from analysts
AI Shift
Full-Stack Builder
AI-Native PM (2026+)
- •Prototype with AI tools
- •Query data in natural language
- •Own revenue outcomes
- •Orchestrate AI agents
The shift is not theoretical. Here are the concrete ways AI has restructured what product managers do every day.
Before (2024)
PM writes PRD, hands to engineering
After (2026)
PM prompts AI to generate working prototype, iterates with engineering on production readiness
Before (2024)
PM requests data pull from analytics team
After (2026)
PM asks Amplitude AI or queries data in natural language, gets instant insights
Before (2024)
PM synthesizes 20 user interviews manually
After (2026)
AI transcribes, tags, themes, and surfaces patterns; PM applies judgment
Before (2024)
PM manages backlog of feature requests
After (2026)
AI auto-triages, deduplicates, and suggests priorities; PM makes strategic calls
Before (2024)
PM coordinates stakeholder updates via meetings
After (2026)
AI generates status summaries from project tools; PM focuses on alignment and decisions
LinkedIn's Full-Stack Builder Experiment
The most dramatic proof that Naval's tweet was describing reality — not predicting it — came from LinkedIn itself. Under CPO Tomer Cohen, LinkedIn did something unprecedented: they killed their APM program and replaced it with the Associate Product Builder (APB) program.
What Changed at LinkedIn
Traditional APM
- ✗Resume-based application
- ✗PM-specific skills only
- ✗Rotational program
- ✗Separate PM, design, eng tracks
- ✗Write specs for others to build
New APB (Associate Product Builder)
- ✓60-second demo of something you built
- ✓PM + design + engineering combined
- ✓Cohort-based, project-driven
- ✓Single "Full-Stack Builder" career ladder
- ✓Build it yourself with AI tools
The application is radical: no resume. Instead, applicants share a 60-second demo of a product they built that people actually use, and answer questions about how they built it using AI tools. As LinkedIn put it: “We want your work, not your job history.”
This is not an experiment at a startup. This is LinkedIn — a Microsoft subsidiary with 1B+ members — making an institutional bet that the PM of the future is a builder, not a document-writer.
The PM-to-Engineer Ratio Flip
Perhaps the most structurally significant claim in this shift comes from Andrew Ng, who suggested the developer-to-PM ratio could flip from the traditional 1:4-6 (one PM for every four to six engineers) to 2:1 (two PMs for every one engineer).
Traditional Model
Bottleneck: Engineering capacity
AI Era Model
Bottleneck: Product judgment
When AI accelerates code generation, the scarce resource shifts from building to deciding what to build. Product judgment becomes the bottleneck.
The logic is straightforward: when AI can generate code at unprecedented speed, the bottleneck shifts from building to deciding what to build. Engineering bandwidth is no longer the scarce resource — product judgment is. Organizations need more people who can identify the right problems, frame the right solutions, and evaluate the right outcomes.
This doesn't mean PM headcount doubles overnight. It means that the skills traditionally associated with product management — customer empathy, problem framing, prioritization, outcome evaluation — become distributed across the organization. Everyone on the team needs more “PM thinking,” even if their title says engineer or designer.
The Vibe Coding Hangover
Not everyone is celebrating. As vibe coding went mainstream, a serious counter-narrative emerged — what some are calling the “Vibe Coding Hangover.”
Security Warning
METR research found that applications built through unreviewed vibe coding were 40% more likely to contain critical security flaws. The speed that makes vibe coding powerful also makes it dangerous when used without proper oversight.
Karpathy himself acknowledged this tension in his one-year retrospective. He noted that when he coined the term, “LLM capability was low enough that you'd mostly use vibe coding for fun throwaway projects, demos and explorations. It was good fun and it almost worked.” But now, “programming via LLM agents is increasingly becoming a default workflow for professionals, except with more oversight and scrutiny.”
This is where PMs become essential. The gap between a vibe-coded demo and a production product is precisely the gap that product management fills: quality standards, security review, user testing, edge case handling, scalability planning, and outcome measurement. AI can build fast. PMs ensure what gets built actually works for users at scale.
The AI PM Tools Ecosystem
AI tools for PMs are no longer experimental. They are standard infrastructure across every phase of the product lifecycle.
| Tool | Category | Use Case |
|---|---|---|
| ChatPRD | Document Generation | AI-generated PRDs, specs, and strategy docs from conversational prompts |
| Amplitude AI | Analytics | Natural language queries on product analytics; automated insight generation |
| BuildBetter | Research Synthesis | AI-powered call analysis, customer feedback synthesis, and research summarization |
| Cursor / Replit Agent | Prototyping | PMs building functional prototypes through conversational AI coding |
| Dovetail AI | User Research | Automated tagging, theming, and pattern detection across user interviews |
| Notion AI | Knowledge Management | Intelligent search, auto-summarization, and writing assistance across team docs |
| Linear + AI | Project Management | Auto-triage, duplicate detection, and intelligent priority suggestions for backlogs |
| Claude / ChatGPT | General Purpose | Strategy brainstorming, competitive analysis, draft writing, data analysis, and code review |
The Key Shift
These tools have moved from “nice-to-have experiments” to “core operating infrastructure.” Research from BuildBetter shows 78% of product teams report more accurate decision-making after adopting AI-based analytics tools, and teams using AI-powered collaboration platforms have cut time-to-market by 25%.
Job Market: Bifurcation in Real Time
The PM job market in early 2026 is a story of two realities playing out simultaneously.
Shrinking: Traditional PM
Booming: AI-Native PM
The Contraction
- ↓6 in 10 companies plan 2026 layoffs
- ↓44% cite AI as a primary factor
- ↓PMs named among top 4 roles cut in AI restructuring
- ↓Traditional “spec-writing” PM roles shrinking fastest
The Expansion
- ↑600+ AI PM roles open (and growing)
- ↑$180-260K+ total compensation
- ↑AI PM: +20-40% salary premium over generalist PM
- ↑“Full-Stack Builder” roles emerging at scale
The pattern is clear: AI is not eliminating PM roles — it is polarizing them. PMs who embrace AI and can demonstrate builder capabilities are commanding premium compensation and abundant opportunities. PMs who position themselves purely as coordinators and document-writers face a shrinking market.
92% of product leaders now own revenue outcomes directly, up from roughly 70% two years ago. The mandate is not just to ship features — it is to drive measurable business impact. AI fluency is table stakes; business outcome ownership is the differentiator.
Implications for PMs: The 12-Week Upskilling Roadmap
Whether you are a senior PM looking to stay relevant or an aspiring PM looking to break in, here is a structured approach to building AI-native PM capabilities.
AI Literacy
Weeks 1-4→ Speak confidently about AI in stakeholder meetings
Hands-On Building
Weeks 5-8→ Portfolio piece: idea → working prototype
AI Product Judgment
Weeks 9-12→ Make informed AI product decisions
Strategic AI PM
Ongoing→ Go-to person for AI product strategy
Phase 1: AI Literacy (Weeks 1-4)
- •Understand how LLMs, fine-tuning, RAG, and AI agents work at a conceptual level
- •Learn the difference between prompting, few-shot learning, and system prompts
- •Read Andrej Karpathy's "Intro to LLMs" talk (available on YouTube)
- •Take Andrew Ng's "AI for Everyone" course on Coursera (free)
Outcome: You can speak confidently about AI capabilities and limitations in stakeholder meetings
Phase 2: Hands-On Building (Weeks 5-8)
- •Build a functional prototype using Cursor, Replit Agent, or Claude Code
- •Create an internal tool for your team using AI coding assistants
- •Use ChatPRD or similar tools to generate your next PRD, then critically edit it
- •Prototype a feature concept end-to-end without involving engineering
Outcome: You have a portfolio piece showing you can go from idea to working prototype
Phase 3: AI Product Judgment (Weeks 9-12)
- •Learn evaluation frameworks for AI outputs (accuracy, hallucination, bias)
- •Design human-in-the-loop workflows for AI-assisted features
- •Understand cost-quality-latency tradeoffs in AI product decisions
- •Study case studies of failed AI products to understand PM failure modes
Outcome: You can make informed decisions about when and how to deploy AI in products
Phase 4: Strategic AI PM (Ongoing)
- •Lead an AI feature from discovery through launch with measurable outcomes
- •Build frameworks for evaluating AI vendor vs. build-in-house decisions
- •Develop expertise in a vertical (AI for healthcare, fintech, enterprise, etc.)
- •Contribute to internal AI governance and responsible use policies
Outcome: You are the go-to person for AI product strategy in your organization
Job Search Strategies for the AI Era
If you are actively looking for PM roles in 2026, the game has changed. Here is how to position yourself.
Build a "Shipped with AI" Portfolio
CriticalCreate 2-3 working prototypes using AI coding tools (Cursor, Replit Agent, Claude Code). Record 60-second demos — LinkedIn's APB application format is becoming an industry standard. Show that you can go from problem to working product.
Reframe Your Resume Around Outcomes + AI
CriticalReplace "managed backlog of 200+ items" with "identified $2.3M revenue opportunity through AI-assisted customer research and delivered 34% conversion improvement." Show you use AI to drive outcomes, not just to be busy.
Target the AI-Expanding Market
HighFocus on companies where AI is creating new PM roles: AI-native startups, enterprise companies standing up AI product teams, and companies building AI infrastructure. The 600+ open AI PM roles offer $180-260K+ TC.
Develop an AI Product POV
HighHave a clear, articulate perspective on AI product strategy. Write 2-3 posts about AI product decisions. Show that you can think critically about where AI adds value and where it doesn't — this is what separates AI-fluent PMs from AI-adjacent ones.
Network in AI-PM Communities
MediumThe intersection of AI and PM is where the energy is. Join communities, contribute to discussions, attend meetups. The people shaping the "Full-Stack Builder" movement are hiring from within these networks.
Growing Within Your Current Role
Already in a PM role? Here is how to leverage this shift for career growth rather than being threatened by it.
Become the AI Champion on Your Team
Introduce one AI tool to your workflow this week. Use ChatPRD for your next PRD, Amplitude AI for your next data analysis, or Claude for competitive research. Share results with your team. The person who brings AI-powered efficiency to the team gets recognized for the team's improved output.
Impact: Immediate visibility + sets the standard for the team
Build Something — Even If It's Not Your Job
Use a weekend to build a working prototype of an idea that has been sitting in your backlog. Use Cursor or Replit Agent. Show it in your next sprint review. The PM who shows a working demo instead of a wireframe changes the conversation entirely.
Impact: Demonstrates builder capability + accelerates product decisions
Own a Revenue-Tied Metric
With 92% of product leaders now owning revenue outcomes, volunteer to own a metric tied to business impact. This is where AI-augmented PMs have massive leverage — you can run more experiments, analyze more data, and iterate faster than ever before.
Impact: Career-defining move for promotion to Senior/Staff PM
Document Your AI-Powered Process
Create an internal playbook for how your team uses AI tools. This positions you as a thought leader internally and creates leverage for a leadership role. Product Ops teams are actively looking for PMs who can define AI-augmented processes.
Impact: Leadership positioning + org-wide influence
Specialize Where AI and Domain Intersect
AI PM specialists command a 20-40% salary premium. Identify where your domain expertise (healthcare, fintech, enterprise, etc.) intersects with AI capabilities. This compound expertise is extremely difficult to automate and extremely valuable.
Impact: Long-term career moat + premium compensation
What Comes Next: Agentic Engineering
In his one-year retrospective, Karpathy himself signaled that vibe coding's moment — at least as originally conceived — is ending. His proposed successor: “agentic engineering.”
The implications for PMs are profound. “Orchestrating agents and acting as oversight” is exactly what product managers do — just applied to AI systems instead of human teams. The PM who can orchestrate AI agents to build, test, and iterate on products — while applying human judgment about what should be built — is the highest-leverage role in the AI era.
Naval's tweet, Karpathy's evolution, LinkedIn's APB program, Andrew Ng's ratio flip — these are not disconnected signals. They are convergent evidence that product management is becoming the core operating layer of AI-native organizations. Not because the old skills don't matter. Because those skills — customer empathy, problem framing, prioritization, outcome evaluation — are exactly what AI systems need most and cannot provide for themselves.
The Bottom Line
AI is not replacing product managers. It is making product management the most important function in the organization. The PMs who recognize this shift — and build accordingly — will not just survive the transition. They will lead it.
Sources & References
- Naval Ravikant (@naval), “Vibe coding is the new product management” — X, Feb 3, 2026
- Andrej Karpathy, original “vibe coding” tweet — X, Feb 3, 2025
- Andrej Karpathy, one-year retrospective & “agentic engineering” — X, Feb 4, 2026
- Aakash Gupta, response thread on PM implications — X, Feb 4, 2026
- Lenny's Newsletter, “Why LinkedIn is replacing PMs with AI-powered full-stack builders” — Tomer Cohen interview
- Product School, “Will AI Replace Product Managers?” — referencing Andrew Ng's ratio analysis
- OfficeChai, “Vibe Coding Is The New Product Management” — analysis
- DEV Community, “What Is Vibe Coding in 2026? One Year From Karpathy's Tweet”
- Stack Overflow 2025 Developer Survey — vibe coding adoption statistics
- BuildBetter.ai Research — AI tool adoption impact on product team decision-making, Feb 2025
- McKinsey Global AI Report 2024 — enterprise AI productivity and ROI data
- Collins Dictionary — 2025 Word of the Year: “vibe coding”
- METR Research — security vulnerability analysis of AI-generated code
Frequently Asked Questions
What did Naval Ravikant mean by "vibe coding is the new product management"?
Naval argued that the core PM skill set — understanding user problems, framing what to build, and evaluating whether the output solves the problem — is exactly what "vibe coding" requires. The person who can articulate the right prompt and judge the output is doing product management, just without the traditional bureaucratic layer. It means PM skills are becoming the highest-leverage capability in AI-native teams.
Will AI replace Product Managers?
No, but it will dramatically reshape the role. AI automates the execution layer (writing PRDs, data analysis, prototyping) while elevating the importance of judgment, strategy, and customer empathy — all core PM skills. McKinsey data shows only 11% of companies have realized measurable ROI from AI at scale, precisely because they lack the product thinking needed to deploy it effectively. PMs who embrace AI will replace PMs who do not.
What is the "Full-Stack Product Builder" role?
Pioneered by LinkedIn's CPO Tomer Cohen, the Full-Stack Product Builder combines PM, design, and engineering into one cross-functional role. LinkedIn replaced its traditional APM program with an Associate Product Builder (APB) program in 2025. Applicants submit a 60-second demo of a product they built (no resume required). This model is spreading across the industry.
How should PMs upskill for AI-native product management?
Start with three layers: (1) AI Literacy — understand how LLMs, fine-tuning, RAG, and agents work at a conceptual level; (2) Hands-on practice — use Claude, ChatGPT, Cursor, or Replit Agent to build prototypes; (3) Product judgment for AI — learn to evaluate AI output quality, identify failure modes, and design human-in-the-loop systems. Stack Overflow reports 84% of developers are already vibe coding. PMs need to match this fluency.
What is the PM-to-engineer ratio shift Andrew Ng described?
Andrew Ng suggested the developer-to-PM ratio could flip from the traditional 1 PM per 4-6 engineers to 2 PMs per 1 engineer. This happens because AI dramatically accelerates code generation, meaning the bottleneck shifts from building to deciding what to build. Product judgment, not engineering bandwidth, becomes the scarce resource.
Is vibe coding safe for production products?
Not without oversight. METR research found applications built through unreviewed vibe coding were 40% more likely to contain critical security flaws. Karpathy himself acknowledged this, proposing "agentic engineering" as the mature successor — where humans orchestrate AI agents with proper oversight, testing, and security review. The gap between a demo and a production product is still massive.
How does this affect PM job searches in 2026?
AI PM roles are surging with 600+ open positions offering $180-260K+ total compensation. But traditional PM roles face pressure — 6 in 10 companies plan 2026 layoffs and 44% cite AI as a factor. Job seekers should: (1) demonstrate AI fluency in portfolios, (2) show prototypes built with AI tools, (3) highlight product judgment and customer empathy as differentiators, and (4) target companies where AI is creating new PM roles rather than eliminating them.
What is "agentic engineering" and why does it matter?
Coined by Andrej Karpathy in February 2026 as the successor to vibe coding, agentic engineering is a professional discipline where developers and PMs orchestrate AI agents rather than writing code directly. "Agentic" because AI agents do the implementation; "engineering" because there is real skill, oversight, and rigor involved. Karpathy noted that vibe coding's original meaning — "forget that the code even exists" — was fine for throwaway projects but insufficient for production. Agentic engineering is the grown-up version.
About the Author
Aditi Chaturvedi
·Founder, Best PM JobsAditi is the founder of Best PM Jobs, helping product managers find their dream roles at top tech companies. With experience in product management and recruiting, she creates resources to help PMs level up their careers.