what is ai coding agent

Michael Macor/The San Francisco Chronicle via Getty Images Earlier this year, former OpenAI exec Andrej Karpathy coined a new term — “vibe coding” — for using artificial intelligence to rapidly develop software using natural language prompts. But the approach comes with some glaring shortcomings that have gradually come to light, from major cybersecurity problems leading to mass leaking of sensitive personal information to rampant hallucinations that turn

OpenAI OpenAI has announced a brand-new AI-powered web browser, dubbed ChatGPT Atlas, which is now available globally. Perhaps most memorably, Atlas features an “agent mode” — initially limited to paying Plus, Pro, and Business users — that acts as a natural extension of OpenAI’s forays into agentic AIs, meaning models intended to autonomously complete tasks. In the context of a browser, that means that users can watch the AI move the cursor as it navigate

Illustration by Tag Hartman-Simkins / Futurism. Source: Getty Images When Amazon Web Services suffered an outage on Monday morning, it practically took down the internet with it. There went all of Amazon’s services, from its shopping hub to its Ring doorbell cameras. ChatGPT went quiet. “Smart” mattresses became unsleepable. Video games like Fortnite blinked out, as did platforms like Snapchat and banking apps. An even greater source of alarm was how long

Getty / Futurism As ChatGPT’s third anniversary approaches, AI hype is hard to escape. Tech companies are committing hundreds of billions of dollars to enormous data center buildouts, while industry leaders are promising an inevitable future in which superintelligent machines will usher in an era of unprecedented prosperity. But many tech insiders are quietly singing a very different tune, expressing a growing conviction that’s been circulating for a while

Apple is not playing around with its lawsuit against OpenAI, its onetime partner and now apparent rival, over allegedly stolen trade secrets. After filing a bombshell lawsuit against the AI company last week, the Cupertino giant has reportedly sent legal letters to roughly 40 former employees who now work at OpenAI. The Financial Times reports, citing unnamed sources, that the letters direct the former employees to preserve relevant documents and communica

Google open-sources the Elephant-Goldfish Model plugin its engineers use to catch AI coding mistakes Google has open-sourced the Elephant-Goldfish Model plugin, an internal tool that makes an AI coding agent write and defend a design document before it touches any code. The plugin hands that document to fresh AI sessions that have never seen the design conversation and tells them to find what is wrong with it. The EGM plugin in a nutshell (Source: Google)

Forgotten AI agents may still access sensitive company data AI agents that have not run recently retain credentials and connections to company systems, according to Arrakis Security’s 2026 Autonomous Agents Threat Report. About 76% of enabled agents examined had not run in 60 days, leaving access available through inactive software. Agents by exposure family (Source: Arrakis Security) The analysis covered thousands of agents in customer production environm

AI coding agents fall for malicious instructions planted in the Downloads folder A file that a website saves to the Downloads folder without approval can trick AI coding agents into installing a malicious npm package or handing over the user’s login token when they’re opened in that folder, Forever Security has found. The file is called AGENTS.md, a plain text document developers place in a project folder with instructions for coding agents, the AI tools t

Getty / Futurism Your pesky remote freelancers demanding more money as inflation soars? You could try replacing them with AI agents instead — but it probably won’t work out well. New research highlighted by Wired shows how these AI models designed to automate tasks — if not entire jobs — turn out to be incredibly unproductive compared to the humans they’re replacing. Conducted by researchers at the nonprofit Center for AI Safety (CAIS) and the massive data

Getty / Futurism Researchers at Google’s Threat Intelligence Group (GTIG) have discovered that hackers are creating malware that can harness the power of large language models (LLMs) to rewrite itself on the fly. An experimental malware family dubbed PROMPTFLUX, identified by GTIG in a recent blog post, can rewrite its own code to avoid detection. It’s an escalation that could make future malware far more difficult to detect, further highlighting growing c

Longtime fears of AI going rogue are starting to feel more real, so much so that some lawmakers now want powerful AI systems to come with an emergency kill switch. U.S. Reps. Ted Lieu, a California Democrat, and Nathaniel Moran, a Texas Republican, introduced a bipartisan bill Thursday that would require major AI companies to make sure they have the ability to throttle, suspend, or shut down their most powerful AI. The so-called AI Kill Switch Act would al

Artificial intelligence has always had the capacity to present a meaningful improvement in accessibility for people with disabilities. Unfortunately, accessibility devices usually aren’t thought to have major commercial applications, so they often get coopted by tech bros who think every innovation is for them. Case in point: the technology of Augmental. The startup pretty explicitly markets itself as an accessibility technology company. On its “About” pag

Framework of GeneAgent for gene-set analysis. Credit: Nature Methods (2025). DOI: 10.1038/s41592-025-02748-6 Researchers at the National Institutes of Health (NIH) have developed an artificial intelligence (AI) agent powered by a large language model (LLM) that creates more accurate and informative descriptions of biological processes and their functions in gene set analysis than current systems. The work is published in the journal Nature Methods. The sys

GPT-6.1 Terra meets OpenAI’s Critical cyber threshold, and it’s cheaper than GPT-5.6 Terra OpenAI released GPT-6.1 Terra, a model that operates desktop software, to all Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and to developers through its API as gpt-6.1-terra. API prices are 50 percent below those of GPT-5.6 Terra and GPT-6.1 Sol: $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. T