Apple just launched Xcode 26.3, turning its primary development environment into an agentic coding hub that can parse entire project structures, edit files, and run multi-step workflows. The release natively plugs in Anthropic's Claude and OpenAI's Codex alongside Model Context Protocol support, keeping developers in control via a strict permission layer.
OpenAI just open-sourced Symphony, an orchestration tool that turns Linear boards into autonomous agent workflows. Instead of supervising coding models line by line, teams assign issues and let Symphony dispatch isolated Codex runs that implement features, pass CI, and record video walkthroughs for human sign-off. It shifts developer focus from watching agents code to managing the actual work.
OpenAI and Figma just rolled out a new Codex-to-Figma integration powered by the Figma MCP Server. Teams can now turn Figma canvases into working code and reverse code back into editable Figma designs across Design, Make, and FigJam. Why it matters: It turns product development into a seamless roundtrip, letting engineers design and designers build without switching contexts.
OpenDataLoader PDF turns messy PDFs into clean Markdown and JSON at over 100 pages per second on a single CPU core, no GPU required. Built specifically for LLM and RAG pipelines, it preserves table structures, strips hidden prompt-injection text, and is completely free for commercial use.
Perplexity just introduced Perplexity Computer, an end-to-end AI system designed to handle complex, multi-step workflows. Instead of just answering search queries, the system unifies current AI capabilities to autonomously research, design, code, deploy, and manage projects from start to finish. It marks a clear shift from conversational search into full-fledged agentic AI.
Nous Research just open-sourced Hermes Agent, an autonomous assistant that runs across Telegram, Discord, and your terminal. It features a built-in learning loop that turns completed tasks into reusable skills and builds a persistent model of you over time. It’s fully MIT-licensed and connects to any LLM provider.
Alex Finn fed Cursor's new 'AI records its own demo video' announcement into his OpenClaw agent Henry, which fully rebuilt the feature and recorded a demo in just five minutes. The setup is now embedded in their workflow, automatically attaching demo clips to every pull request.
Anthropic just introduced Remote Control for Claude Code, letting developers manage terminal tasks straight from their smartphones. You can kick off a command on your local machine and take over from the Claude mobile app or browser while stepping away for a walk or meeting. Claude continues running in the background, freeing engineers from their desks.
OpenAI just made its specialized coding model, GPT-5.3-Codex, available to all developers through the Responses API. The model combines frontier coding performance with deep professional knowledge in a single package. It basically gives builders immediate programmatic access to integrate its top-tier code generation directly into their own apps.
Google Labs just introduced an agent step to Opal, upgrading its workflow builder from static model calls to dynamic agency. The agent autonomously selects the right tools and models (like Search or Veo) for the job, remembers preferences across sessions, dynamically routes logic, and can stop to ask follow-up questions when it needs clarification. It gives builders the flexibility of an agent without losing control over the pipeline.
Cursor just launched cloud agents that run in isolated virtual machines, giving them dedicated environments to build, interact with, and test software in parallel. Instead of handing off unverified diffs, each agent iterates independently and returns merge-ready PRs complete with video demos and logs. Cursor says over 30% of its own merged PRs are now built autonomously by these agents.
Confluence Labs just exited stealth with a 97.9% score on the ARC-AGI-2 benchmark at $11.77 per task, open-sourcing its entire solver on GitHub. The lab uses LLM-guided program synthesis to tackle data-sparse reasoning, aiming to bring the approach to drug design and hardware engineering next.
Google just restricted access to its Antigravity vibe coding platform for developers using the open-source agent OpenClaw, with some even reporting banned Google accounts. Google says users were siphoning Gemini tokens through third-party platforms outside its terms, degrading performance for others. The crackdown comes just a week after OpenClaw's creator joined OpenAI, prompting the tool to drop Google support entirely.
Anthropic just announced that Claude Code can automate legacy COBOL modernization, shrinking multi-year mainframe overhaul projects down to quarters. The tool maps code dependencies and automates analysis that previously required massive consulting teams, dealing an immediate blow to traditional IT service providers and triggering a 10% plunge in IBM shares.
Docker just detailed how to run OpenClaw, the open-source AI coding agent, inside isolated micro VM sandboxes. By pairing it with Docker Model Runner, developers can run local models completely offline without cloud fees or risking leaked API keys.
Anthropic just outlined how its agentic tool Claude Code can tackle the multi-decade backlog of COBOL modernization. The ancient language still handles roughly 95% of US ATM transactions, but original developers are long gone. AI automates dependency mapping, workflow documentation, and test drafting, squeezing modernization timelines from years down to quarters while engineers steer business logic.
OpenAI just declared SWE-bench Verified unreliable for evaluating frontier coding models and stopped reporting its scores. An audit revealed that 59.4% of problems failed by OpenAI o3 had flawed test cases that rejected valid solutions, while multiple frontier models memorized human patches from training data. Why it matters: When models pass tests by reciting memorized training data, benchmarks stop measuring real-world coding capability.
OpenAI just boosted the inference speed of GPT-5.3-Codex-Spark by roughly 30%, pushing throughput past 1,200 tokens per second. That blistering output makes real-time code generation and autonomous developer workflows feel virtually instantaneous. OpenAI says broader speed upgrades across its model lineup are coming next.
Anthropic just rolled out new developer updates to Claude Code on desktop. The tool can now preview live running applications, review code, and handle CI test failures and PRs in the background. Why it matters: It removes routine friction from development by letting Claude quietly fix broken builds without interrupting your flow.
Anthropic just launched Claude Code Security in preview, bringing automated vulnerability hunting and patching straight into Claude Code on the web. Powered by Claude Opus 4.6, the system reasons through code context like a human auditor instead of relying on basic rule-matching, having already uncovered over 500 long-standing bugs in open-source projects. For now, it's open to Enterprise and Team tiers, with free access for open-source maintainers.
