Artificial intelligence continues to polarize opinion, with stakeholders viewing the technology as either an invaluable professional asset, a glorified search engine, or a vastly overrated development. This lack of consensus persists as industry leaders promise a new industrial revolution, while skeptics dismiss the advancements as marketing noise and some former executives issue stark warnings regarding safety.
Public discourse around the impact of AI intensified recently following a viral essay by investor and former CEO Matt Shumer, who argued that the technology is poised to automate virtually any computer-based job. Shumer detailed how he utilizes AI to manage application builds, claiming the system can handle user flow, design, and tens of thousands of lines of code. He further posited that if AI can demonstrate such proficiency in coding, it may soon gain the ability to autonomously iterate and improve its own outputs.
However, Shumer’s claims face scrutiny from researchers, particularly following his 2024 acknowledgment of exaggerating AI performance—a move he described as the biggest mistake of his professional life. While Shumer maintains he uses OpenAI’s GPT-5.3 Codex tool for complex testing, critics remain skeptical of the feasibility of his claims given the lack of specific, verifiable details regarding the models employed.
Much of the public disagreement stems from the significant gap between free AI services and premium, subscription-based tools. Matt Murphy, a partner at Menlo Ventures, notes that there is a broad spectrum of user exposure to AI, which is evolving rapidly. Paid tiers frequently offer advanced agent capabilities that move beyond basic chatbots, allowing for more complex work management. For example, Anthropic’s Claude Cowork agent and OpenAI’s Codex are largely confined to subscription models costing $20 per month or more.
According to Carnegie Mellon University professor Emily DeJeu, assessing AI based solely on free services is misguided, as these versions fail to showcase the full technical potential of the software. Oren Etzioni, professor emeritus at the University of Washington, characterizes this divide as the difference between a novice intern and a seasoned professional. While free tools can provide sophisticated dialogue, they fall short when tasked with high-level legal or paralegal responsibilities.
The distinction between tiers is further blurred as companies incorporate more advanced features into free versions. Anthropic recently introduced its Sonnet 4.6 model, intended to bring performance closer to its top-tier Opus models. Stanford University’s James Landay argues that such integration suggests the gap between free and paid AI may be narrowing, potentially lessening the perceived divide in capabilities.
Skepticism remains regarding the pace of AI adoption and its true efficacy. Recent studies have tempered the narrative surrounding AI productivity; for instance, research from the Center for AI Safety and Scale AI revealed that prominent models struggled with data visualization and game development tasks. Furthermore, a July study by Model Evaluation and Threat Research found that developers actually required 19% more time to complete coding tasks when using AI tools, based on data from early 2025.
Landay contends that the focus on AI in software engineering is inherently biased, as code follows a logical, structured format that is uniquely suited for machine processing. He emphasizes that many professional roles lack such structured requirements, meaning AI’s success in coding should not be extrapolated to other industries. Concerns about the broad automation of knowledge work were nonetheless exacerbated in early February when Anthropic released industry-specific AI helpers, causing a temporary decline in software stock valuations. The report also notes that after a viral essay from an AI CEO and investor claimed the tech is coming for any job that involves sitting in front of a compu, the discrepancy in how people view AI has perhaps never been so apparent as this past week. The report also notes that plus fewer limits on usage, but those who pay get access to another feature: Agents that can handle some work for you rather than just chatbots that craft responses. The report also notes that for example, is only available in the $20-per-month Pro plan and higher, anthropic’s Claude Cowork agent. The report also notes that and no one seems to agree on which is right. The report also notes that yet it’s all being referred to in the same way, but there may be a simpler explanation as to why people have taken such divergent stands: People use different types of AI in different ways. The report also notes that “I’ll tell the AI: ‘I want to build this app. The report also notes that here’s roughly what it should look like, here’s what it should do.
