Michael Truell, chief executive of AI coding startup Cursor, has warned that “vibe coding”, a growing trend where developers rely heily on AI-generated code without fully understanding it, could create weak foundations in software systems. He said this while speaking at a Fortune Brainstorm AI conference. Truell said while AI tools can help developers move faster, using them without strong engineering discipline may lead to fragile code that is hard to maintain, scale or secure in the long run.
Speaking about the rapid adoption of generative AI coding assistants, Truell said the problem is not the technology itself but how it is being used. He also explained that some developers accept AI-written code simply because it appears to work, without properly reviewing its structure, logic and long-term impact. This approach may se time initially but can cause serious technical debt later.
What is ‘vibe coding’ and why it worries expertsAdd WION as a Preferred Source
“vibe coding” is commonly used to describe a casual style of development where programmers rely on AI suggestions, prompts and auto-generated solutions instead of carefully designing systems. Truell warned that this mindset risks turning software engineering into guesswork, especially when teams stop asking why a piece of code works and focus only on whether it runs.
He said software built this way may function at first but can fail under pressure, such as during scaling, security testing or integration with other systems. Over time, fixing such issues becomes more expensive and complex. Truell added that strong fundamentals, such as understanding system design, testing and documentation, remain essential even in an AI-assisted world.
AI tools need human judgement, not blind trustTrending Stories
Top 9 air-to-air missiles used by today’s most advanced fighter jets
9 air-to-air missiles that give modern fighter jets first-strike advantage
‘Deport Every Indian’: American journalist calls for mass deportation and threatens violent attack against Indians
WATCH: Man attempts suicide at Mecca’s Masjid al-Haram, dramatically sed by security officer
Mirage, Rafale, Eurofighter and more: Which aircraft can launch the SCALP missile?
Top 10 precision-guided missiles that modern fighter jets rely on
World’s 10 best food cities of 2025-26: Italian cities occupy 6 spots; Indian city bags 5th spot
Christmas apocalypse has been cancelled, says Ghana 'prophet', buys Mercedes with donations
NASA’s James Webb telescope captured supernova from the dawn of universe
Iron Beam, HELIOS, DragonFire and more: 6 most advanced laser weapons and where India stands
‘US Ny’s first operational carrier-based U’: What is Boeing and Rolls-Royce powered MQ-25 Stingray?
New Friday OTT Releases (December 26, 2025): Stranger Things 5, Cashero, Revolver Rita- 5 New movies & shows releasing on Netflix, ZEE5 and more
Epstein Files: Trump witnessed the killing of my newborn baby, 13-year-old girl claimed
Epstein files: Mohammed Bin Salman, Pope John paul II, IDF sweater, massage manual and other sensitive photosTruell made it clear that Cursor itself is designed to help developers write better code, not replace careful thinking. He said AI tools should act as assistants that speed up work, while humans remain responsible for decisions, reviews and architecture. He stressed that developers must still read, test and understand what AI produces, especially for core systems.
Industry experts quoted in similar reports he echoed this concern, warning that unchecked AI-generated code could introduce security flaws and reliability issues. Truell said companies that balance speed with discipline are more likely to benefit from AI in the long term, while those chasing quick results may face problems later.
A warning as AI coding tools grow fastThe warning comes at a time when AI coding assistants like Cursor, GitHub Copilot and others are seeing rapid adoption across startups and large companies, the rise of these tools has changed how developers work, but leaders like Truell believe the basics of software engineering cannot be ignored.