AI Coding Live News

CrowdStrike reported on Oct 29 that the China-nexus espionage group VAPOR PANDA, starting Oct 17, ran the open-source agentic intrusion framework ARTEX with DeepSeek plus a second, unconfirmed commercial LLM, hitting more than 200 organizations in at least 14 countries from a single controller over 11 days. Post-exploitation was confirmed at nine organizations, where the BLOODALCHEMY backdoor was deployed. Planning agents used about 1.4 billion input tokens in a single intrusion, and recovered transcripts show a human operator steering the agents in Chinese at key decision points.

LLM agents operating in novel API or terminal environments frequently repeat identical errors due to an absence of procedural memory. Existing memory bootstrapping approaches rely heavily on curated prompt playbooks or oracle verifiers that require extensive domain foresight. Illuin Technology releases DAEDALUS (arXiv:2610.08048), an autonomous dual-agent framework bootstrapping reusable memory entirely from self-generated practice without human oracles. DAEDALUS pairs an Explorer agent—which probes the execution environment to formulate curriculum-style tasks—with a Solver agent. When the solver stumbles, it extracts candidate heuristics that are committed into a persistent memory bank only after proving their efficacy across repeated verification trials. Across AppWorld, tau^2-bench, and AutomationBench, DAEDALUS boosts average success rates by up to 15.9 percentage points and raises Pass^5 rates by 2.2x over memory-less baselines, with generated procedural memories transferring seamlessly across heterogeneous foundation model families.

Microsoft Agent Framework now ships a first-class ClaudeAgent over the Claude Agent SDK (pip install agent-framework-claude --pre), sharing the same BaseAgent surface as Azure OpenAI/Copilot agents so Claude drops into SequentialBuilder workflows. Built-in Read/Write/Bash/Glob, custom tools, MCP, permission modes, and streaming.

_akhaliq shared HarnessDev, a paper asking whether LLMs can create and evolve their own agent harness—the runtime scaffolding for tools, loops, and execution—instead of relying only on hand-written agent frameworks. It matters for coding and research agents that need cheaper, self-improving harnesses.

DeepSeek-V4-Flash-Vision-Exp is live on the API: it matches V4-Flash on text—including agents, reasoning, and world knowledge—while jumping on multimodal agent benchmarks to near Claude Opus 4.8. A free Files API and DeepSeek Harness 0.1.1 ship with it; images bill at V4-Flash rates (up to 384 tokens each) and drop into existing agent frameworks.

All-Hands AI announced OpenHands 0.18, bringing architectural upgrades to its open-source software engineering agent. With a redesigned action planner, dynamic rollback search, and multi-agent code verification, OpenHands achieved a 53.4% solve rate on SWE-bench Verified, establishing a new SOTA among open-source agent frameworks.

Google DeepMind upgraded its coding agent framework to Antigravity 2.5. The update introduces an in-memory asynchronous event bus that decouples Planner, Coder, and Auditor agents into parallel actor loops, increasing throughput across complex distributed project builds by 70%.

Microsoft's .NET team announced production stability for the GitHub Copilot coding harness combined with Microsoft Agent Framework across .NET and Python, featuring MCP tools, sandboxes, and approval workflows.