惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
V
Vulnerabilities – Threatpost
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
V
Visual Studio Blog
月光博客
月光博客
IT之家
IT之家
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
T
Tailwind CSS Blog
罗磊的独立博客
S
SegmentFault 最新的问题
博客园 - 三生石上(FineUI控件)
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
量子位
V
V2EX
Jina AI
Jina AI
The GitHub Blog
The GitHub Blog
小众软件
小众软件
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
阮一峰的网络日志
阮一峰的网络日志
Recent Announcements
Recent Announcements
MongoDB | Blog
MongoDB | Blog
Y
Y Combinator Blog
H
Help Net Security
博客园_首页
Cyberwarzone
Cyberwarzone
T
Tenable Blog
A
Arctic Wolf
C
CERT Recently Published Vulnerability Notes
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
T
Threat Research - Cisco Blogs
aimingoo的专栏
aimingoo的专栏
Google DeepMind News
Google DeepMind News
博客园 - 叶小钗
C
Cyber Attacks, Cyber Crime and Cyber Security
美团技术团队
Attack and Defense Labs
Attack and Defense Labs
GbyAI
GbyAI
博客园 - 【当耐特】
Cloudbric
Cloudbric
NISL@THU
NISL@THU
B
Blog RSS Feed
K
Kaspersky official blog
Hugging Face - Blog
Hugging Face - Blog
P
Privacy International News Feed
博客园 - Franky
博客园 - 司徒正美
Microsoft Azure Blog
Microsoft Azure Blog
Apple Machine Learning Research
Apple Machine Learning Research
Webroot Blog
Webroot Blog
Microsoft Security Blog
Microsoft Security Blog

Moor Insights & Strategy

RESEARCH NOTE: Computex 2026 Shows How Infrastructure Fragments as AI Scales Is SAP's AI Transformation the Future of SaaS? - Pulse Brief OpenAI Flexes Enterprise Ambitions With Colin Fleming As Business CMO RESEARCH NOTE: Google I/O 2026 — More Details on AI and AR Glasses, Including Project Aura BROADCAST ANALYSIS: Patrick Moorhead Discusses the AI Market, Semiconductors, SpaceX, and Big IPOs on The Street, June 10, 2026 At Cisco Live 2026, Cisco Bets The Network Is The AI Platform MI&S Weekly Analyst Insights — Week Ending June 5, 2026 Apple WWDC 2026 - Resetting Siri, OS Improvements, and Parental Controls BROADCAST ANALYSIS: Patrick Moorhead Discusses NVIDIA Computex, China Trade Restrictions, and Berkshire’s Google Investment on CNBC Asia, June 1, 2026 RESEARCH NOTE: Dell Makes Its Case for Owning the Enterprise AI Stack Microsoft Work Trend Index 2026 Shows AI Productivity Is Not Enough Huawei's Chip Claims, SpaceX IPO Insights, Network X, Starcloud, AT&T & Amazon Leo Updates RESEARCH NOTE: Can Intel Wildcat Lake Challenge Apple’s MacBook Neo and Make Cheap PCs Great Again? ANALYST INSIGHT: Tenstorrent Is Disrupting the Inference Market MI&S Weekly Analyst Insights — Week Ending May 29, 2026 RESEARCH NOTE: Panasonic TOUGHBOOK 56 Brings Much-Needed Updates to the Rugged Form Factor RESEARCH NOTE: Amazon’s Acquisition of Globalstar Accelerates Amazon Leo Ambitions RESEARCH NOTE: IBM Turns Sovereignty Into a Product ANALYST INSIGHT: Mission-Critical ERP Needs Mission-Critical Agents RESEARCH NOTE: Cadence Leans into EDA Super Agents at Cadence LIVE 2026 MI&S Weekly Analyst Insights — Week Ending May 22, 2026 RESEARCH NOTE: Distance Technologies Partners on Kia Vision Meta Turismo Concept Car Retail AI Requires a Fundamentally Different Approach to Implementation — Research Brief BROADCAST ANALYSIS: Patrick Moorhead Discusses NVIDIA Earnings on CNBC, May 20, 2026 Enterprises Need To Be Careful Before They Go All-In On Anthropic RESEARCH NOTE: AT&T, T-Mobile, and Verizon Create Unprecedented Joint Venture for D2D Satellite Simplicity MI&S Weekly Analyst Insights — Week Ending May 15, 2026 Carriers Form D2D Satellite JV, 6G Expectations Cool & Data Center Pushback in Socorro RESEARCH NOTE: Google’s Gemini Enterprise Agent Platform Is a Serious Bid for the Agentic Control Plane BROADCAST ANALYSIS: Patrick Moorhead Discusses NVIDIA and U.S.–China Trade Relations on CNBC, May 13, 2026 RESEARCH NOTE: Motorola’s All-New Razr Fold Headlines a Mostly Unchanged Razr Lineup RESEARCH NOTE: SAP’s Bet on an Open Data Foundation for Agentic AI RESEARCH NOTE: Samsung Galaxy S26 Ultra — Samsung’s Halo Is Better Than Ever MI&S Weekly Analyst Insights — Week Ending May 8, 2026 Nvidia & Corning Unite, NTIA Report, ConnectX, FWA Uplink and 6G Spectrum News RESEARCH NOTE: Adobe CX Enterprise, An Agentic Control Plane for Orchestrated Customer Experience and AI Discovery RESEARCH NOTE: T-Mobile’s New SuperBroadband Aims to Solve Business Broadband Pain Points BROADCAST ANALYSIS: Patrick Moorhead Discusses AMD Earnings and Arm on CNBC, May 6, 2026 RESEARCH NOTE: Samsung’s Redesigned Galaxy Book6 Pro with Intel Core Ultra 3 Is a Welcome Upgrade RESEARCH PAPER: From Devices to the Cloud — Arm's Relevance in the Age of AI RESEARCH NOTE: Qlik’s Bet on Production-Grade Agentic AI RESEARCH NOTE: Google TPU 8: Architecture, Context, and Enterprise Relevance ANALYST INSIGHT: How Google’s Agentic Data Cloud Redefines What Context Means for the Enterprise MI&S Weekly Analyst Insights — Week Ending May 1, 2026 T-Mobile Super Broadband, Fiber Expansion, Satellite MVNO Rumors, & Big Tech Earnings — The 6G Podcast RESEARCH BRIEF: Oracle's Blueprint for Agentic AI RESEARCH NOTE: Devices Launched at MWC 2026 — Smartphones, Robots, AI, and PCs BROADCAST ANALYSIS: Patrick Moorhead Discusses Hyperscaler Earnings on CNBC, April 29, 2026 ANALYST INSIGHT: Google Cloud’s AI Hypercomputer at Next 2026: Real Co-Design, Targeted Reach RESEARCH NOTE: Meta Ray-Ban Display: Bridging the Gap Between Smart Glasses and AR AI Canvases Move From Collaboration To Core Revenue And IT Operations RESEARCH NOTE: Samsung Galaxy XR Headset: A Strong Hardware Foundation Waiting on Software DataCenter Podcast: Episode 58 — We’re Talking AI Bottlenecks, Google Cloud Next TPU 8 Review MI&S Weekly Analyst Insights — Week Ending April 24, 2026 RESEARCH NOTE: First-Take Analysis: Nuvacore Emerges From Stealth Mode RESEARCH NOTE: The HP Z2 Mini G1a: A Tiny Powerhouse for the AI Workstation Era RESEARCH NOTE: HP Imagine 2026: HP Evolves in the Era of AI BROADCAST ANALYSIS: Patrick Moorhead Discusses Apple's New CEO and Future Strategic Direction on CNBC, April 20, 2026 RESEARCH NOTE: Lenovo Closes Infinidat Acquisition — What Does It Mean for Enterprise Storage? MI&S Weekly Analyst Insights — Week Ending April 17, 2026 Amazon’s Globalstar Deal, Verizon’s FIFA Play, and Millimeter Wave Insights — The 6G Podcast RESEARCH NOTE: Galileo Brings Cisco a Purpose-Built Agent Evaluation Layer RESEARCH NOTE: Cohesity Positions AI Resilience as the Foundation for Enterprise AI Adoption DataCenter Podcast: Episode 57 — We’re Talking Beyond the Border, Nutanix .NEXT Recap RESEARCH NOTE: The HP EliteBoard G1a: A Capable PC in an Innovative Form Factor RESEARCH NOTE: Samsung’s Galaxy S26 Lineup Leads with AI and Privacy RESEARCH NOTE: Velaura AI’s Titan Core Targets the Biggest Problem in AI Datacenter Silicon: Power RESEARCH NOTE: The ASUS ROG Xbox Ally X Has Rekindled My Hope for Windows Gaming Handhelds RESEARCH NOTE: Infor Positions Industry Context as the Foundation for Agentic ERP BROADCAST ANALYSIS: Patrick Moorhead Discusses Advanced Chip Packaging on CNBC, April 8, 2026 PULSE BRIEF: Navigating Supply Chain Constraints with Architectural Flexibility RESEARCH NOTE: MWC 2026 Showcases Semiconductors for 5G, 6G, and Many Kinds of AI RESEARCH BRIEF: From Infrastructure to Resilience Foundation — Reframing Cyber Resilience for Data Management PULSE BRIEF: Cloud-Native Edge AI Platforms RESEARCH PAPER: The Economic Impact of a Domestic Semiconductor Foundry RESEARCH NOTE: Arm Enters the Silicon Business with AGI CPU RESEARCH NOTE: The Inference Inflection Point: What NVIDIA’s Groq 3 LPX Really Signals for Enterprise AI BROADCAST ANALYSIS: Patrick Moorhead Discusses Arm AGI CPU on CNBC, March 25, 2026 DataCenter Podcast: Episode 56 — Artificial “Stupidity” and Arm Enters the AI Race PULSE BRIEF: Density Is Destiny — Rethinking AI Infrastructure in the AI Data Era BROADCAST ANALYSIS: Patrick Moorhead Discusses Arm's New AGI CPU on CNBC, March 24, 2026 BROADCAST ANALYSIS: Patrick Moorhead Discusses NVIDIA GTC Announcements on CNBC, March 16, 2026 RESEARCH NOTE: WD Innovation Day and FY2026 Q2 Earnings Reflect Disciplined Execution RESEARCH NOTE: AWS and Cerebras Partner to Deliver Disaggregated AI Inference The Enterprise Applications Podcast, Ep 26: AI Agents - The New Control Layer for Enterprise Apps DataCenter Podcast: Episode 55 — The AI Power Problem: Data Centers, Nuclear SMRs, and AWS + Cerebras RESEARCH NOTE: VAST Forward 2026 Positions the Data Platform as the Persistent Operational Layer for AI Game Time Tech Ep 28: MLB 2026 Season – AI, XR, Stadium Tech, and the Future of Baseball BROADCAST ANALYSIS: Patrick Moorhead Discusses AI Chip Export Controls and Oracle's Upcoming Earnings on Yahoo Finance, March 9, 2026 RESEARCH NOTE: Digging into the AMD–Meta Deal RESEARCH NOTE: Zoom Promotes ‘System of Action’ via AI-First Canvases and Agentic Workflows Game Time Tech Ep 27: How AI Is Transforming Pro Sports RESEARCH NOTE: IBM FlashSystem — Advancing Toward an Intent-Aware Storage Control Layer The Enterprise Applications Podcast - Ep 25: Is Enterprise ERP Ready for Agentic AI? RESEARCH NOTE: RPT-1 Is Turning SAP Data Into Insightful AI RESEARCH NOTE: Dell Pro 14 Premium Laptop with 5G Connectivity BROADCAST ANALYSIS: Patrick Moorhead Discusses NVIDIA Earnings on Yahoo Finance, February 25, 2026
RESEARCH NOTE: Rayfin Turns Microsoft Fabric Into a Runtime for Agent-Built Apps
Mike Leone · 2026-06-13 · via Moor Insights & Strategy

Coding agents have made the front end of software almost trivial to generate, while standing up a secure, governed production backend has barely gotten easier. At its Build 2026 conference, Microsoft introduced Rayfin, an open-source toolset that lets developers and their coding agents define a complete application backend in code and deploy it onto Microsoft Fabric. The move turns Fabric from a destination for analytics into a governed runtime for the applications now built by agents, closing the distance between where enterprise data lives and where it gets acted on.

Where Agent-Built Apps Tend to Stall

The fastest-growing move in software development right now is describing an application in plain language and watching an agent build it. GitHub Copilot has passed 20 million users, and natural-language tools like Replit have turned code generation from autocomplete into something close to delegation. The front of the development cycle has collapsed from weeks into an afternoon, and producing a working app no longer requires an engineer.

What agents haven’t made easier is everything that comes after the prototype works. A generated app still needs a database, authentication, access policies, storage, and safe plumbing, and in an enterprise all of that has to connect to systems that the security and compliance teams already trust. That’s where most agent-built apps stall, and the deeper problem comes with the data those apps touch. Enterprises have spent years consolidating their information into governed data estates, where the most sensitive data sits under strict access and audit controls. No security team will let a quickly assembled app carrying customer records run outside that estate, so the applications that matter most are the ones least able to use the fast new path.

Closing that gap calls for a backend an agent can define as easily as it builds the front end. The definition has to live in code as one authoritative source, so the database, its API, and the app’s own code never drift apart the way generated code usually does. The access rules belong in that same code rather than a separate dashboard, and deployment has to land the finished app inside the governed estate the enterprise already runs, so that each new app strengthens that estate instead of spawning another silo.

How Microsoft Rayfin Builds a Governed Backend in Code

Rayfin, now in public preview, is Microsoft’s attempt to make building a secure production backend as fast as generating the app it serves. It’s an open-source SDK and command-line tool, available on GitHub, that lets a developer or their coding agent declare an application’s backend in code. Data models, business logic, APIs, identity, and access policies live together in one code definition, and from it Rayfin generates the database, the API, and the application code that calls them, so those pieces share a single source of truth instead of falling out of sync. The security rules sit in the same file as the data, so a row-level policy lives next to the column it protects rather than in a separate dashboard. Deploying lands the whole backend on Microsoft Fabric, where it runs as a fully managed part of the platform. It signs users in through Microsoft Entra, the company’s identity service, and stores its data in OneLake, Fabric’s data lake. The data inherits Fabric’s governance and compliance by default, usable right away by Power BI, notebooks, and Fabric’s data agents.

Access splits by audience. A developer installs Rayfin and works in an editor such as VS Code, while a business user reaches the same backend through Replit, the natural-language platform Microsoft has partnered with, which an administrator can wire to deploy straight into the company’s Fabric tenant. Both the developer and the business user stay where they already work without sidestepping governance. Replit’s chief executive framed the division of labor plainly, saying that agents write the code and Fabric ships it quickly and safely. A presenter from tool maker Leatherman, an early customer running both Fabric and Replit, described his team finally developing fast using its preferred tools while keeping applications on governed enterprise data.

Rayfin is one part of Microsoft’s broader push to make Fabric a production backend for applications, alongside other Build launches including Azure HorizonDB, a Postgres-based database, and Fabric IQ (now in GA) as a shared context layer for agents. Fabric brings real scale to that role, now past 31,000 customer organizations and running at more than $2 billion in annual revenue — which means it has grown 60% over the past year.

Governance Wins with the CIO, but the Developer Ultimately Decides

Some have likened Rayfin to a Heroku for AI-native applications, which holds in spirit, though it’s more opinionated and built for agents to target rather than humans. Speed at the front of the build stopped being the constraint a while ago. The scarce thing now is getting a generated app safely into production, and that is the work Rayfin compresses. What turns the compression into a purchase is governance, because absorbing agent speed into a controlled environment is the part almost no one has solved. Call it assurance — the confidence that a new app touches the right data, enforces the right rules, and can be reconstructed for an auditor later. Every app a team builds through Rayfin inherits that assurance the moment it ships, picking up Entra identity, OneLake data, and Fabric’s policies at deployment time instead of after a security review weeks later.

At scale, that approach should bring shadow AI under control the same way platforms once tamed shadow IT, leaving a CIO one identity model, one data layer, and one set of policies rather than a patchwork per project. The packaging is a crucial point of differentiation, with the security rules sitting in the same file as the data and a Replit on-ramp for non-developers who would otherwise build outside the tenant entirely.

The harder question is whether developers will actually use it. Governance is what a CIO wants, sure, but to a developer that same governance can feel like being locked in. Microsoft could sell Rayfin to the CIO and still watch the people who write the code skip it, because they can always build the app their usual way and add the controls later. So which apps actually use Rayfin will show whether the pitch worked. If it mostly draws small new projects rather than the important apps that hold sensitive customer data, then the governance pitch hasn’t really landed, and Fabric ends up as a place to host apps more than the place people build them. Rayfin is also still in preview, and the idea of defining a backend in code is one that other vendors are already chasing, so Rayfin will be judged on whether it holds up at real scale, rather than simply being first.

It’s worth flagging that Rayfin ties an app closely to Fabric, although that closeness is also where the value comes from. The governance is automatic precisely because the integration runs deep, so the biggest payoff naturally comes for teams already on Fabric. Microsoft is widening the path from there, keeping the SDK open source and signaling support for other places to run it over time. Microsoft’s platform breadth puts it among the front-running hyperscalers, pulling analytics, databases, a semantic layer, and now app development onto one platform while most point-solution rivals go deep in just one. The other hyperscalers are chasing the same broad vision, which keeps this a real race. What matters next is making it easy to start small, smoothing the transition away from older and competing systems, and answering the questions that Build left open about pricing, availability, and how Rayfin overlaps with Power Apps, Microsoft’s existing low-code app builder.

Owning the Data Estate and the Software on Top

Enterprise software is shifting from a craft of writing code to a discipline of governing what gets built. Once anyone can stand up a working application by describing it, the bottleneck stops being how fast the thing gets made and becomes whether anyone can vouch for what it does. Boards and regulators are already asking which data a given app reached and who authorized it, and those questions get harder when an agent built the app. The organizations pulling ahead treat that scrutiny as a reason to centralize what their people build rather than a reason to slow them down.

Rayfin is Microsoft’s bet that whoever owns the governed data estate is best positioned to own the application layer that agents are building on top of it. The natural home for that bet is the software that runs inside the business, where stitching together separate data, AI, and app tools gets expensive once audit and governance enter the picture — and where a single governed platform can win on cost and control. That’s where Rayfin is aimed, with consumer-scale apps left to the best-of-breed combinations that already fit them. Within that home, the platform that already holds the data can now host the software that runs against it. Whether the bet pays off depends on the months after the preview, as Microsoft works to turn demos into customers running Rayfin in production, extend the open-source promise past Fabric’s own walls, and win over the developers who decide whether Fabric becomes a place to build or only a place to deploy.