This post describes five engineering and algorithmic interventions developed at Microsoft to stabilize reinforcement learning post-training of multimodal agents for Copilot at production scale.

Engineering@Microsoft

performance

Cloudflare

231 min read

Frontend•2026-02-24

How we rebuilt Next.js with AI in one week

One engineer and an AI model rebuilt Next.js from scratch as "vinext", a Vite-based drop-in replacement deployable to Cloudflare Workers, for ~$1,100 in tokens over one week.

Code Mode: give agents an entire API in 1,000 tokens

Cloudflare introduces Code Mode, a technique that exposes their entire 2,500+ endpoint API to AI agents via just two MCP tools consuming only ~1,000 tokens.

Architecture•2026-02-19

Our Multi-Agent Architecture for Smarter Advertising

Spotify introduces a multi-agent AI architecture to unify advertising media planning across Direct, Self-Serve, and Programmatic buying channels.

Machine Learning•2026-02-12

How low-bit inference enables efficient AI

This article explores low-bit inference techniques that make large AI models faster and more cost-efficient to serve in production.

Introducing Markdown for Agents

Cloudflare introduces Markdown for Agents, a feature that automatically converts HTML pages to markdown at the network level for AI crawlers and agents.

How we built the Microsoft Learn MCP Server

Microsoft built a remote MCP Server to expose Microsoft Learn documentation to AI agents via the Model Context Protocol standard.

Engineering@Microsoft

Insights from our executive roundtable on AI and engineering productivity

Dropbox hosted an executive roundtable on AI and engineering productivity, sharing lessons from their own AI adoption journey and cross-industry leadership perspectives.

Culture

developer productivity

2025 Q4 DDoS threat report: A record-setting 31.4 Tbps attack caps a year of massive DDoS assaults

Cloudflare's Q4 2025 DDoS Threat Report documents a record-breaking 31.4 Tbps attack and a year of unprecedented DDoS escalation.

Google’s AI advantage: why crawler separation is the only path to a fair Internet

This post argues that Google's dual use of its search crawler (Googlebot) for both indexing and generative AI training gives it an unfair competitive advantage over publishers and rival AI companies.

Introducing Moltworker: a self-hosted personal AI agent, minus the minis

This post introduces Moltworker, a middleware solution that enables running Moltbot (a self-hosted personal AI agent) on Cloudflare's infrastructure without dedicated hardware.

Cursor at Grab: Adoption and impact

This post details how Grab scaled adoption of the Cursor AI coding assistant across engineering and non-technical teams, sharing key metrics and productivity findings.

Engineering

Microsoft

01 min read

Machine Learning•2026-01-28

Diagnosing instability in production-scale agent reinforcement learning

This post identifies a late-phase instability mechanism in production-scale reinforcement learning for tool-using agents, caused by tool-conditioned variance amplification.

Engineering@Microsoft

Cloudflare

131 min read

Frontend•2026-01-16

Astro is joining Cloudflare

Cloudflare has acquired The Astro Technology Company, the team behind the Astro web framework, committing to its continued open-source development.

From deployment slop to production reality: How BriX bridges the gap with enterprise-grade AI infrastructure

This article introduces BriX, a platform that transforms AI prototypes into enterprise-grade production tools without requiring deep DevOps expertise.

Human Native is joining Cloudflare

Cloudflare has acquired Human Native, a UK-based AI data marketplace that transforms multimedia content into licensed, structured training data.

Machine Learning•2025-12-18

Inside the feature store powering real-time AI in Dropbox Dash

Dropbox Dash built a custom hybrid feature store to power real-time AI ranking across tens of thousands of work documents.

Background Coding Agents: Predictable Results Through Strong Feedback Loops (Honk, Part 3)

Spotify shares how they designed reliable background coding agents ("Honk") using strong verification loops to minimize incorrect or broken pull requests at scale.

The Interaction Changes Everything: Treating AI Agents as Collaborators, Not Automation

This post presents a framework for treating AI agents as collaborative partners rather than automation tools, based on Microsoft's Entra SDK v1-to-v2 migration across hundreds of repositories.

Engineering@Microsoft

AI agent

Developer productivity