best ai coding agent 2026

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)

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

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

Ransomware gang talked AI agents into doing much of its dirty work The group behind the Aur0ra ransomware has been using AI agents to move through victim networks, collect credentials and work its way up to domain administrator access, according to Sophos. AI-assisted intrusion workflow (Source: Sophos) The company, which tracks the group as GOLD OGRE, recovered artifacts from the Cursor AI coding assistant in late August that show the attackers handing th

Meta’s financial metrics have taken a notable punch due to its AI spending. In the company’s earnings report on Wednesday, Meta executives shared that free cash flow for the past quarter was down to only $784 million, a major drop compared to the $8.55 billion number for the same time period last year. Revenue for the entire fiscal year is also looking likely to come below market expectations. Meta is the second AI hyperscaler in a week to receive investor

Claude’s Duet agent is a little neurotic and surprisingly helpfulDuet is a decent helper, if a little jumpy at times.Let’s get to work. | Photo: Allison Johnson / The Verge is a senior reviewer with over a decade of experience writing about consumer tech. She has a special interest in mobile photography and telecom. Previously, she worked at DPReview.Normally, the products I test don’t offer to help me write up my experience with them. Duet is different.Cl

Illustration by Tag Hartman-Simkins / Futurism. Source: Getty Images You kind of have to sympathize with some of these so-called AI agents. They’re apologetic screw-ups that can’t get anything right, while their cruel human masters insist on making them carry out important work anyway. Inevitably this goes wrong, and the AI model must beg for forgiveness. The latest example of this comes from a user on Reddit who claims they were using Google’s Antigravity

Alibaba just introduced its most powerful AI model to date, sending a fresh shiver down the spine of every American tech leader who fears their Chinese counterparts gaining the upper hand. The model, called Qwen3.8-Max, comes with 2.4 trillion parameters and specializes in long-horizon agentic tasks, extending over as long as several days and “with minimal human involvement,” as Alibaba put it in its announcement. In one internal test, the model worked aut

Your pinned MCP server may not be as pinned as you think An MCP server configured to run one fixed release on an administrator’s workstation ran with 15 different sets of dependencies over five weeks, with its configuration untouched. James Lagermann, Director of Technical Alliances at Corelight, found the drift in his company’s lab. Twenty-three of the server’s 69 packages changed versions, including the cryptography library, the JSON Web Token library an

An average of 8 minutes is all hackers need to log in with stolen credentials Attackers used stolen credentials to log in an average of eight minutes after infection across 27 confirmed infostealer cases, according to a report from Elastic Security Threat Command. Researchers also documented misuse of AI applications to execute commands and digitally signed remote monitoring tools to maintain access. “An agent that holds credentials and can reach the inter

Security teams want to spot unusual AI agent behavior while they’re still defining normal AI agents topped the list of planned investments in behavioral monitoring in Abnormal AI’s 2026 State of Behavioral Security Report. Asked where their organizations would invest most over the next two years, 43% of respondents chose AI agents, more than three times the share who chose API integrations, the next most common answer. Security professionals were divided o

Microsoft researchers demonstrate hidden backdoor technique for AI agents Microsoft researchers have demonstrated how small changes to an AI model can implant a hidden backdoor, causing an agent to execute an attacker’s command when a specific trigger appears. The technique, called Lazarus, requires no model retraining and allows the agent to behave normally during standard safety evaluations. The company has not observed threat actors weaponizing the tech

Who collects the keys when an AI agent retires? Think about the last time a good engineer left your company. Someone collected the laptop and the badge. IT revoked their logins before the end of the day. Their manager reassigned the open tickets, and HR filed their records somewhere they’ll stay for years, in case anyone ever challenges a decision they made. Now think about the last time an AI agent left. Someone probably deleted it from the dashboard. I’v

OpenAI’s Daybreak brings AI agents to security testing and vulnerability fixes OpenAI is bringing AI agents into more of the software security process with Daybreak, pairing web application penetration testing with tools that investigate vulnerabilities and develop fixes. The two offerings, Red Team and Defense Factory, connect the work of finding security weaknesses with patching and retesting them. Red Team tests applications and verifies fixes Daybreak

The rise of large language models (LLMs) has equipped AI agents with the ability to interact with users through natural, human-like conversations. As a result, these agents now face dual responsibilities: engaging in dialogue and performing complex reasoning or planning tasks. Their conversational interactions must be contextually informed, while their actions need to work toward achieving specific goals. This balance between conversation and multi-step re