The Information · Tech & AI
TIER 5 Tue, 21 Oct 2025 17:32:50 +0000 (UTC)
Around the start of this year, executives at artificial intelligence providers like OpenAI and Salesforce said 2025 would be “the year of agents,” making it seem like fully autonomous AI would soon take over human jobs. Since then, Salesforce, Microsoft, SAP and many others pitched roughly seven types of agent to automate white collar tasks in realms like coding, HR, finance, and sales. But the promise of truly autonomous agents still hasn’t been fulfilled, and tech leaders are warning companies to lower their expectations. Andrej Karpathy, one of the founding members of OpenAI and a leading voice in AI research, believes it’ll take at least a decade until AI can meaningfully automate entire jobs, “like an employee or intern that you would hire to work with you.” In a podcast interview, he gave a brutal assessment of AI’s current capabilities, in part because large language models lack cognition, and argued that most jobs still can’t be automated. | | | | Oct 21, 2025 ---|--- | # Applied AI --- | | By Aaron Holmes ---|--- | | | | Supported by | ---|--- | | _Hi! If you’re finding value in our Applied AI newsletter, I encourage you to consider subscribing to The Information. It contains exclusive reporting on the most important stories in tech, like this story from Catherine on_how AI is impacting sites’ search traffic _but, paradoxically, not their revenue.Save up to $250 on your first year of access._ * * * Welcome back!Around the start of this year, executives at artificial intelligence providers like OpenAI and Salesforce said 2025 would be “the year of agents,” making it seem like fully autonomous AI would soon take over human jobs. Since then, Salesforce, Microsoft, SAP and many others pitched roughly seven types of agent to automate white collar tasks in realms like coding, HR, finance, and sales. But the promise of truly autonomous agents still hasn’t been fulfilled, and tech leaders are warning companies to lower their expectations.Andrej Karpathy, one of the founding members of OpenAI and a leading voice in AI research, believes it’ll take at least a decade until AI can meaningfully automate entire jobs, “like an employee or intern that you would hire to work with you.” In a podcast interview, he gave a brutal assessment of AI’s current capabilities, in part because large language models lack cognition, and argued that most jobs still can’t be automated.“Overall, the models are not there,” Karpathy told podcaster Dwarkesh Patel. “I feel like the industry is making too big of a jump and is trying to pretend like this is amazing, and it’s not. It’s slop. They’re not coming to terms with it, and maybe they’re trying to fundraise or something like that.”Karpathy’s remarks resonated with business leaders on the front lines of using AI agents, which have repeatedly run into speed bumps over the past year when trying to get AI to handle complex workplace tasks without errors.For instance, French online document storage firm AODocs has been using AI coding tools such as Microsoft’s GitHub Copilot and Anthropic’s Claude Code for software development, but none of the tools are reliable enough for the company to cut headcount or fully replace jobs, said CEO Stéphan Donzé. AODocs also sells AI features powered by models from OpenAI and others that let customers automate discrete tasks like checking whether a job application has the right files attached, but Donzé says he’s been careful not to oversell the features because they still require human oversight.“It‘s the fashion of the day to call these tools ‘agents,’ but agent doesn’t really mean autonomous,” Donzé said. “There‘s the marketing BS that says everything will be autonomous, and then there’s some level of truth behind it, but they can only be autonomous in limited use cases.”Others say they’re using AI agents to replace some employees, but note that the agents need to be supervised more than those employees typically would. Sahil Lavingia, CEO of the e-commerce marketplace app Gumroad, said he’s shrunk the size of his engineering team to roughly a dozen people, down from over 40 two years ago, thanks to AI coding agents, but said remaining staff spend a lot of their time double checking the agents’ work.“I think it’s interesting that I can replace some $400,000-a-year jobs [with agents], but I can’t completely replace them. I just have someone else I pay $400,000 to manage agents to become four times as productive and then fire three people,” Lavingia said. “AI-generated code needs to be [checked] as the product experience can’t be validated by the AI’s yet, and that’s a different task from coding.”And not all agents are created equal. For instance, Karpathy acknowledged that so far AI excels the most at coding tasks, but noted that it has a harder time with tasks without a clear right or wrong answer that models can be trained on, such as generating slide decks, which he said is “much, much harder” for that reason. I’ve consistently heard the same from engineers at Microsoft, for instance, who have been trying to wrangle AI features in its Office software to generate Powerpoint documents more reliably.Besides coding, the primary enterprise use case for AI agents thus far has been customer support agents meant to automatically handle phone calls or messages from customers and answer their questions or resolve their issues. Karpathy said he thinks that type of role is ripe for automation because call center employees follow predictable scripts, but still warned that current automation of call centers is overhyped.“Maybe [call centers] are swapping in AI, but then I would still wait for a year or two because I would potentially expect them to pull back and rehire some of the people,” he said. “Then I would expect that across the economy. A lot of jobs are a lot harder than a call center employee.”**GPT-5 ‘Finds’ Solutions to Hard Math Problems—But Doesn’t Solve Them** Last week, OpenAI researcher Sebastien Bubeck took to X to share an impressive accomplishment by OpenAI’s GPT-5 model: the AI helped researchers find the solution to several difficult conjectures in mathematics, known as Erdos problems, which were first made decades ago by the 20th century mathematician Paul Erdos. Kevin Weil, vice president of OpenAI for Science, touted the same accomplishment, saying GPT-5 “found solutions” to ten Erdos problems that had “been open for decades.”Some took the announcements as proof that OpenAI’s models were getting better at complex math. But Oxford researcher Thomas Bloom, who maintains the online database of Erdos problems that Bubeck was referring to, deflated some of that enthusiasm on Friday.That’s because, as Bloom pointed out on X, GPT-5 didn’t actually solve the math problems in question; it merely surfaced online publications by other mathematicians who had already solved those problems. The problems were listed as “open” on Bloom’s site because Bloom himself wasn’t aware that others had solved them, he said in a response to Weil.Weil and Bubeck deleted their posts and said in subsequent posts that they meant to clarify that GPT-5 was proving useful at turning up relevant research, not solving math problems. Some of OpenAI’s rivals were quick to pounce, such as Google DeepMind CEO Demis Hassabis, who called Weil’s original post “embarrassing.” While the episode doesn’t prove that GPT-5 is cracking novel math problems, OpenAI can take comfort in the model’s ability to surface relevant research on esoteric topics. To some academics and researchers, that might be worth spending $200 a month for ChatGPT’s most expensive tier. | | A message from Google Cloud --- ### Are you making the most of AI right now? Join Google experts in our learning series to discover how to build an AI-fluent organization, use the latest in agentic AI, and drive measurable business value. Google experts will lead a 3-part masterclass series to help you identify the most impactful AI use cases and build an AI-fluent organization today. **Register now.** ## Upcoming Events #### Tuesday, October 28 – Wednesday, October 29 — The Information’s 2025 WTF Summit The rules are changing, fast. AI, volatile markets, and political uncertainty are reshaping business. Join the boldest women at The Information’s WTF 2025 Summit in Napa Valley. Limited tickets remain. More details | --- _Thank you for reading the Applied AI newsletter! I’d love your feedback, ideas and tips:_aaron@theinformation.com _._ _If you think someone else might enjoy this newsletter, please pass it forward or they can sign uphere._ | | ## Opportunities #### Group subscriptions Empower your teams to stay ahead of market trends with the most trusted tech journalism. Learn more | --- #### Brand partnerships Reach The Information's influential audience with your message. Connect with our team --- | | ## About Applied AI A new The Information franchise that will take you inside how businesses are using AI to automate all kinds of work. Read the archives --- | | Follow us | | | | | | | | | ---|---|---|---|---|---|---|---|--- Sent to registered2nd@gmail.com | Manage your preferences or unsubscribe | Help The Information * 251 Rhode Island Street, Suite 107, San Francisco, CA 94103