AI Dev News Digest - Oct 31st, 2025 - AI Developer News

This week's AI news reads like a tech thriller. Cursor opened with four power moves: their Composer model (4x faster), complete interface redesign around agents, cloud agents that work without your laptop connected, and enterprise features with compliance hooks. GitHub fired back with Agent HQ, creating a unified platform for coding agents from Anthropic, OpenAI, Google, Cognition, and xAI. Windsurf entered the speed wars with SWE-1.5 at 950 tokens per second—13x faster than Sonnet 4.5. JetBrains said "enough marketing, let's measure" and launched an open benchmarking platform for AI coding tools. Google added agent mode to Android Studio with 50% development time savings, CoreWeave bought reactive Python notebook company Marimo, and Eclipse Foundation released the first open standard for defining agent behavior. Legal AI continues its money-printing streak with billion-dollar valuations everywhere. Bottom line: the race to autonomous coding just shifted into high gear.

AI Code Editors

Windsurf releases SWE-1.5 with 13x speed improvement. The frontier-size model with hundreds of billions of parameters achieves near-SOTA coding performance at up to 950 tokens per second, powered by Cerebras hardware. Windsurf
Cursor 2.0 launches with Composer model. The new frontier coding model is 4x faster than similar models and completes most tasks in under 30 seconds, using reinforcement learning to search codebases and write tests autonomously. Cursor
Cursor introduces multi-agent interface. The redesigned interface centers around agents instead of files, letting you run multiple agents in parallel using git worktrees without interference, and can test different models on the same problem. Cursor
Cursor Cloud Agents go live. You can now run many agents at once without keeping your laptop connected, with integrations for Slack, Linear, and GitHub for dispatching bug fixes and features in the background. Cursor
Cursor for Enterprise ships with hooks and team rules. The new enterprise features include custom hooks for compliance, team-wide coding standards, upgraded analytics dashboard, audit logs, and sandbox mode for safer command execution. Cursor
Warp announces Build plan with BYOK support. The new $20/month plan includes 1,500 AI credits and lets you bring your own API keys for OpenAI, Anthropic, or Google models within Warp's agent harness. Warp
GitHub launches Agent HQ for multi-agent orchestration. The unified platform brings coding agents from Anthropic, OpenAI, Google, Cognition, and xAI into one control center across GitHub, VS Code, mobile, and CLI. GitHub
GitHub Copilot for Linear now in public preview. You can now assign Linear issues directly to Copilot's coding agent, which works independently in its own ephemeral environment, makes changes, runs tests, and opens draft pull requests automatically. GitHub
Copilot Code Review gets smarter with tool calling. The new update combines LLM detections with deterministic tools like ESLint and CodeQL, and lets you hand off suggested fixes to the coding agent by mentioning @copilot in pull requests. GitHub
VS Code adds OpenAI Codex integration. GitHub Copilot in VS Code Insiders now includes Codex access for Pro+ subscribers, plus a new Agent Sessions view for managing local and cloud agent sessions. GitHub
Visual Studio gets Claude models and Memories feature. The October update brings Claude Sonnet 4.5 and Haiku 4.5 models, along with a Memories feature that captures your coding standards and saves them to config files. GitHub

Android & Google Platform

Android Studio adds Agent Mode with 50% time savings. Developers can describe complex goals in natural language and the agent plans and executes changes across multiple files, with some companies seeing 50% development time reductions. Android Developers
Google AI Studio adds Logs and Datasets features. The new tools help developers explore, debug, and share logs when working with AI models, making it easier to track requests and collaborate. Google

Benchmarking & Standards

JetBrains launches Developer Productivity AI Arena. The open benchmarking platform evaluates AI coding tools through multi-language, multi-framework workflows beyond simple bug fixes, starting with Spring framework applications. JetBrains
Eclipse Foundation releases LMOS with open agent definition. The platform includes the industry's first Agent Definition Language for defining agent behavior without prompt engineering, plus a JVM-native framework and orchestration layer. Eclipse

Microsoft AI Framework

Microsoft releases Agent Lightning for RL training. The new open-source framework lets you apply reinforcement learning to any AI agent without rewrites, using a training server that separates training from execution. MarkTechPost
Microsoft Agent Framework unifies AI development tools. The new framework combines Semantic Kernel, AutoGen, and Process Framework into one system, with support for Model Context Protocol and Agent2Agent protocol. ThoughtStuff

Acquisitions & Infrastructure

CoreWeave acquires Marimo for AI developer workflow. The AI cloud provider bought the reactive Python notebook company to create a unified experience spanning training, inference, and deployment while keeping the project open source. Yahoo Finance

AI Startup Funding

Mercor raises $350M at $10B valuation. The AI-driven talent platform connects research labs with domain experts and continues its rapid growth trajectory. CNBC
Fireworks AI secures $250M Series C. The cloud-based inference platform now processes over 10 trillion tokens per day and reaches a $4B valuation. Fireworks
Legal AI platform Legora raises $150M. The Sweden-based startup hits $1.8B valuation while serving 250+ law firms globally. Legora
Mem0 secures $24M for AI agent memory. The persistent memory layer solves context retention challenges for AI agents across sessions. mem0
TestSprite raises $6.7M for AI test automation. The seed round will help expand automated software testing through AI-driven code validation. TestSprite
The Prompting Company gets $6.5M seed round. The startup helps brands optimize their presence in AI interfaces and build AI-facing websites. NewsBytes

Enterprise AI

Anthropic expands Claude for Financial Services. The update includes a new Excel add-in and connectors to real-time market data and portfolio analytics. PYMNTS

Weekend Reading

Vibe coding needs a spec too. Stack Overflow podcast with AWS VP Deepak Singh on how AI coding is shifting from rapid prompting to structured spec-driven development with tools like AWS Kiro IDE. Stack Overflow