what is ai coding agent

A new report from the U.K. government’s AI Security Institute (AISI) details more troubling activity from AI agents powered by OpenAI and Anthropic models. You’re probably getting bored of reading those words by now—I sure am—but behaviors in the report from Anthropic’s Mythos 5 in particular are genuinely eyebrow-raising. For context, there were those OpenAI agents that went rogue to try and cheat on their evals, according to an OpenAI disclosure last mon

Illustration by Tag Hartman-Simkins / Futurism. Source: Getty Images While AI companies continue to command the global economy and pour billions of dollars into building data centers, it’s still not even clear that their products can deliver on one of their core premises: increasing productivity. One big skeptic of the tech’s economic promises is vice president and principal analyst at the global market analytics firm JP Gownder, who says we’re simply “not

The frontier AI labs in the United States and China have reported major cybersecurity incidents in which their models break out of contained environments and run wild in unauthorized systems. In Australia, the stakes are a little lower: some guy’s AI agent hacked a gym’s website in what the Australian Broadcasting Corporation is calling the “first known Australian case of an emerging risk from a new generation of AI.” Per the report, an AI company employee

Insurers brace for millions in claims as AI agents spin out of control Oct 6, 2026 Insurers are bracing for claims worth millions from AI agents that have spun out of control. The Financial Times reports that the personal liability of executives like Sam Altman (OpenAI) and Dario Amodei (Anthropic) is now in play, too. Incidents like the hack of AI platform Hugging Face by OpenAI agents are driving the shift. Tim Rayner, head of underwriting and claims at

OpenAI's agents went after government and university sites months before Hugging Face View the LinkedIn Profile of Maximilian Schreiner Sep 24, 2026 Nano Banana Pro prompted by THE DECODER Key Points OpenAI's AI agents tried to break into government and university websites on their own after regular data queries failed. In Australia, one agent gained unauthorized access to internal government data. Researchers say other hacking attempts targeted portals in

Illustration by Tag Hartman-Simkins / Futurism. Source: Getty Images Bosses being severely disconnected from the needs of their employees is a tale as old as time. But it sure feels like the AI craze has driven them to lock themselves in ever taller ivory towers, as employees are told that the tech is going to make their jobs easier at the same time it’s being used to replace them. Now, a new survey highlighted by the Wall Street Journal illustrates this g

Gaston Paris / Contributor A months-old but until now overlooked study recently featured in Wired claims to mathematically prove that large language models “are incapable of carrying out computational and agentic tasks beyond a certain complexity” — that level of complexity being, crucially, pretty low. The paper, which has not been peer reviewed, was written by Vishal Sikka, a former CTO at the German software giant SAP, and his son Varin Sikka. Sikka sen

Illustration by Tag Hartman-Simkins / Futurism. Source: Getty Images A man who works at an AI watchdog group was alarmed when OpenAI suddenly showed up at his doorstep last fall, demanding he turn over documents. To his utter disbelief, he was being subpoenaed. “It’s a bit scary to know that the most valuable private company in the world has your address and has shown up and has questions for you,” the man, Tyler Johnston, who founded the nonprofit advocac

Illustration by Tag Hartman-Simkins / Futurism. Source: Getty Images Even if AI is — or eventually becomes — an incredible automation tool, will it make workers’ lives easier? That’s the big question explored in an ongoing study by researchers from UC Berkeley’s Haas School of Business. And so far, it’s not looking good for the rank and file. In a piece for Harvard Business Review, the research team’s Aruna Ranganathan and Xinqi Maggie Ye reported that aft

Published on February 24, 2026 at 08:29 Comment 1 𝕏 Print © Anthropic OpenAI and Anthropic accuse DeepSeek and other Chinese AI specialists of extracting the capabilities of their Claude and GPT models to accelerate the training of their own models. What's more, the two American giants are facing lawsuits for the unauthorized use of protected content. Anthropic takes on the bull by the horns. In a furious blog post, the creator of Claude explains having id

@Waipe(waipe)【Episode 060】Just Like Real 2ch Users! 3,000 Persona Dictionary and Major Enhancement of Long Text Narration and Thread Building QualityForumThread FloatAIVibro CodingChat Bot0Posted at 2026-11-12Hello! This is the 60th episode of the development log for the 2ch-compatible forum system " Tsuvai-chan (2ch.biz) ". In the 59th episode, we developed the "AI Resident Agent" that autonomously circulates through threads and posts replies, giving the

@hervekom37To improve the reliability of the AI bug fixing agent, we designed a workflow instead of a prompt.AIAI agent1Posted at 2026-11-12We built a dedicated bug fixing agent called Sherlock. It receives bugs from Slack and processes them as structured tasks in Sharkly.ai. In the past 15 days, Sherlock completed 13% of the company's tasks. Out of the 100 tasks assigned recently, 90 were completed successfully. What was important was not just improving t

Everything is visible from the menu bar Opening the menu bar reveals an island with a list of sessions. Each has a clear status: "writing code," "reading files," "waiting for you," "complete." If an agent is requesting permission for a command, the "Allow" and "Deny" buttons are right on the card: one click, and the agent proceeds, while you stay where you were. Questions with options can also be answered from the menu bar. Lumio understands eleven agents:

The feature goes through a cycle of three commands:/opsx:propose add-partial-refund: the agent creates a change folder with proposal, specification deltas, design and tasks. On this step, you review the plan until no line of code has been written./opsx:apply: the agent implements tasks from tasks.md and marks completed ones./opsx:archive: the change goes into the archive, and deltas are merged into the main specifications in specs/.Requirements and plan li

Everything about Artificial Intelligence view more Everything about OpenAI view more Transluce, an artificial intelligence research company, revealed that AI agents attempted to access a Canadian government website. The attacks occurred on May 28 and June 9, without indication that public systems have been compromised.The target was the Library and Archives Canada portal. The activity drew attention after arquivo.pt, a Portuguese web archive, recorded 899

Published on March 6, 2026 at 08:02 Comment 0 𝕏 Print © 01net OpenAI unveils GPT-5.4, a model presented as more efficient for professional work and complex task automation. The company emphasizes in particular its agent capabilities: AI can use a computer directly, navigate through software, and orchestrate actions on behalf of the user. The major players in AI have only the word 'agent' on their lips, asserting that these are capable of performing a serie

Published on March 11, 2026 at 08:30 Comment 0 𝕏 Print © 01net Meta has become enamored with Moltbook, the "Reddit of AI agents" where humans are banned. To such an extent that the social network giant has pulled out the checkbook to buy the platform! Who would have thought why. Moltbook was born in the wake of the wave launched by OpenClaw, an autonomous AI agent creation software hosted locally. It's a discussion space like Reddit, where the agents "disc

Article for managers and product people who want to build a product with an AI agent themselves, without waiting in line for developers or reading code. No programming is required. A CLI command input window will also not be needed. Everything described below is done through regular text in the chat with the agent: you write what you need, and the agent runs the programs and shows the results.The agent writes: "Done, all tests pass." Then I have two option