AI is both your most useful coworker, a glorified search engine or vastly overrated relying on who you ask.
And no one appears to agree on which is correct.
Tech executives championing AI have lengthy spun the narrative that the tech will revolutionize jobs and convey a few new industrial revolution. Skeptics suppose it’s all advertising and marketing hype, whereas some researchers and executives are sounding the alarm about security issues on their approach out the door.
The discrepancy in how individuals view AI has maybe by no means been so obvious as this previous week, after a viral essay from an AI CEO and investor claimed the tech is coming for any job that includes sitting in entrance of a pc.
But there could also be an easier rationalization as to why individuals have taken such divergent stands: People use several types of AI in several methods, but it’s all being referred to in the identical approach.
“There’s just a wide spectrum of how much people have been exposed to the technology, how much they’ve used the technology,” mentioned Matt Murphy, a associate at Menlo Ventures who has led investments in AI corporations together with Anthropic. “And that’s also changing pretty rapidly.”
People who use free AI for primary duties like making grocery lists and planning holidays are probably solely seeing one facet of the know-how. A report from Menlo Ventures printed final June estimated that solely 3% of AI customers are paid subscribers, though Murphy advised NCS he expects that to vary shortly.
But those that pay get entry to a different function: Agents that can deal with some be just right for you moderately than simply chatbots that craft responses, plus fewer limits on utilization.
Anthropic’s Claude Cowork agent, for instance, is just out there within the $20-per-month Pro plan and better. The case is comparable for OpenAI’s Codex coding agent.
It’s that sort of AI that’s fueling issues about AI’s impression on jobs, together with the controversial argument Matt Shumer, an investor and CEO of AI writing assistant startup Otherside AI, will get at in his viral essay.
“I’ll tell the AI: ‘I want to build this app. Here’s what it should do, here’s roughly what it should look like. Figure out the user flow, the design, all of it.’ And it does. It writes tens of thousands of lines of code,” Shumer wrote.
He went on to say that the AI was capable of take a look at the app and make selections regarding style and judgment. And he surmised that if AI may write code that nicely, it may start enhancing upon itself, as nicely.
(AI researchers accused Shumer of exaggerating the efficiency of his firm’s AI mannequin in 2024. He apologized on the time and advised NCS it was the “biggest mistake” of his “professional life” and that he realized by way of the method).
Some specialists are skeptical the use instances Shumer outlined are potential even with paid plans, particularly since he was imprecise about which mannequin he used and what sort of app the AI constructed for him. Shumer advised NCS he primarily makes use of OpenAI’s GPT-5.3 Codex software and that he was engaged on a “medium to high complexity app” for testing functions.
Still, the free model of AI apps doesn’t paint the total image of what the know-how is able to, in response to Emily DeJeu, a professor who teaches programs about using AI in enterprise at Carnegie Mellon University. She mentioned it could be “misguided” to make assumptions about AI’s capabilities primarily based solely on free AI providers.
Oren Etzioni, professor emeritus on the University of Washington and former CEO of the Allen Institute for Artificial Intelligence, described the hole between the free and paid tiers of AI like evaluating an keen but inexperienced intern with a seasoned, hard-working intern. Free AI tiers are good at writing summaries and producing content material, however customers will usually must pay to conduct deep analysis or draft subtle paperwork utilizing AI.
While free AI can “give you surprisingly good advice” and “engage with you in a surprisingly sophisticated dialog, you wouldn’t want to use one of those as your attorney or even as your paralegal,” he mentioned.
But AI corporations are more and more trickling extra superior options all the way down to the free tier, which is a part of why James Landay, cofounder of the Stanford Institute for Human Centered AI, mentioned he doesn’t see a giant distinction between free and paywalled AI. Case in level: Anthropic launched a brand new mannequin referred to as Sonnet 4.6 on Tuesday that it says will carry efficiency nearer to its extra superior Opus fashions solely out there in its paid plans.
Simmering tensions over AI and work
Software shares plummeted in early February after AI firm Anthropic launched a software tailoring its AI helper particularly for particular person industries, like authorized and monetary evaluation. That launch, adopted by Shumer’s essay, stoked issues that AI will finally broadly automate information work the best way it’s beginning to streamline software program engineering jobs.
Yet there’s additionally rising skepticism about whether or not AI resides as much as these lofty declarations, typically made by tech executives with monetary pursuits within the know-how’s success. Some research have poured chilly water over how succesful AI actually is and the way shortly it’s being adopted.
A group of researchers from the Center for AI Safety and Scale AI discovered final 12 months that main AI fashions produced flawed outcomes when tasked with work assignments like visualizing information and coding video video games. An group that exams AI fashions referred to as Model Evaluation and Threat Research present in July that builders take 19% longer to work on their code when utilizing AI, though that analysis was primarily based on instruments from early 2025.
Landay additionally says the position AI is enjoying in software program improvement is overstated within the essay. AI is a useful software that programmers use to hurry up improvement, but it surely’s nonetheless susceptible to errors, and AI fashions aren’t writing themselves. While specialists have broadly agreed AI will change many industries, AI’s proficiency at coding shouldn’t be taken as an indication that it’ll carry out the identical approach in different professions.
“(Coding is) also a logical structure, which is a really good fit for a machine to also be able to test the code and see if it works,” he mentioned. “Many people’s jobs are not structured in that way.”