In a stunning pivot within the tech landscape, Andrej Karpathy, the revered figure behind the term “vibe coding,” has declared this once-trendy method obsolete, raising eyebrows and challenges among founders and developers alike. With his recent transition to Anthropic to foster a new research paradigm dubbed “agentic engineering,” the echoes of past methodologies fade as he heralds a more sophisticated future. This shift not only signals a departure from vibe coding but also embodies a fundamental transformation in how technology breeds creativity and efficiency.
What Changed in the Last Year?
Vibe coding was initially borne out of a desire for accessibility, allowing users to instruct AI models using plain English. It effectively transformed complex coding tasks into manageable projects, with fast outputs—landing pages built in mere hours and prototypes in a weekend. However, the advent of agentic engineering signals a monumental leap forward. Under this framework, AI systems do not merely interpret commands; they autonomously handle code writing, testing, error corrections, and iterative improvements, effectively reshaping the human role from builder to director.
Bridging Old and New Paradigms
Karpathy has framed vibe coding as a transitional phase. Its limitations have become apparent as companies engage with the more robust capabilities of agentic engineering. In interviews with El-Balad, senior engineers from various firms confirm that vibe coding has essentially been abandoned in favor of models that offer deeper automation and require less human intervention. This revelation underscores a growing consensus: companies no longer need to rely on imperfect, piecemeal solutions.
| Stakeholder | Before Vibe Coding | With Vibe Coding | Shift to Agentic Engineering |
|---|---|---|---|
| Founders & Teams | Heavily dependent on developers | Increased speed, but limited scalability | Full-fledged automation; focus on vision instead of execution |
| Regulators | Slow adaptation to tech changes | Struggled with approval processes | Quicker timelines and approvals based on stand-alone AI outputs |
| Developers | Necessary for all tasks | Use of simple tools for quicker outputs | More specialized roles, overseeing AI productivity |
The Ripple Effect Across Global Markets
As the conversation surrounding vibe coding dissipates, the reverberations are felt across multiple markets including the US, UK, CA, and AU. In the US, a neobank recently leveraged agentic engineering to launch a customer-facing feature in four hours, a feat that would have taken months a year ago. This evolution signifies a shift in compliance frameworks adapting to rapid AI capabilities, offering quicker responses from regulators.
Across the UK and CA, the embrace of such technologies hints at a competitive edge in sectors where time-to-market is crucial. Similarly, in AU, many startups are seeking to integrate AI-driven solutions into their compliance strategies, fundamentally altering how financial products are developed and delivered.
Projected Outcomes: What to Watch
As we brace for the next evolution in technology spurred by agentic engineering, three developments deserve close attention:
- Acceleration of Feature Delivery: Expect cycle times for feature development to shrink significantly, influencing hiring and resource allocation trends.
- Shift in Skills Demand: The equilibrium of technical skills will tilt towards project management, as founders focus on directing agents rather than developing bugs.
- Regulatory Adaptation: The pace with which regulations will adapt to AI-engineered developments may shift, impacting approvals and governance models globally.
Karpathy’s return to the forefront with agentic engineering is not merely a trend; it reflects a deeper transformation in tech and entrepreneurship. As we enter this uncharted territory, being ready for a future where shipping outputs are possible without extensive coding knowledge marks a critical evolution in the fabric of innovation.
