Instance Master @InstanceMaster

Small but kinda nice: the #MCP Server template now ships bundled in the .NET SDK. No more installing a separate package. Just: dotnet new mcpserver -o MyMcpServer It shows up in dotnet new list out of the box, and its servicing rides along with the web stack. github.com/dotnet/core/bl…

Tomorrow I'll be at the Small Business Expo in South San Francisco. Going to swap notes with Bay Area SMB owners on what AI and modern software can actually do for them. If you run a small business and have a process you've been meaning to automate, come find me. Free to attend. thesmallbusinessexpo.com/city/san-franc…

My book Web Development with Blazor is currently #1 in the #ASPNET category on Amazon 🎉 Fourth edition. Updated for .NET 10 and C# 14. Foreword by @danroth27 . And yes, probably the cutest raccoon so far. Thank you, everyone ❤️ #Blazor #dotnet

Why move #Blazor WebAssembly off Mono and onto CoreCLR? Mono runs in the browser, but it does not share the optimization investment CoreCLR gets on the server. CoreCLR on WASM means faster execution, real JIT, and one runtime from server to browser. Same code, same behavior, everywhere. linkedin.com/posts/yuriy-fr…

Most of the cleanup work I do isn't bad code. It's the right thing never getting defined in the first place. That's the part worth screening a vendor for.

devInstance 🇺🇸 @DevInstance

2 days ago

A lot of the projects that land on our desk are someone's second try. The first vendor was cheap, moved fast, demoed well, and then left behind software nobody could maintain. By the time the owner comes to us, they are gun-shy and asking a sharper question than they asked the

0 0 1 23 0

Maybe you built it yourself with AI coding tools. Maybe a cheap vendor moved fast, showed great demos, then disappeared. Either way, you now have software that mostly works until it doesn't, breaks in ways nobody can explain, and is too risky to change. I stabilize it, make it maintainable, and turn it into something you can actually build on. devinstance.net/services/code-…

Happy Friday !!!

Four numbers that don't match, one company: Your CRM says 1,200 customers. Accounting says 1,140. The marketing platform has 1,310. Support has 980. Which one is right? Trick question. None of them, because they were never built to agree. Disconnected systems don't just waste time on duplicate data entry. They quietly erode trust in your own numbers, until every important decision starts with an argument about whose spreadsheet is correct. Integration fixes the root cause. When systems sync in real time with validation in between, there's one answer instead of four. Sales reflects inventory. Accounting reflects payroll. The data stops contradicting itself. We map your current systems, find the high-impact connection points, and build integrations that hold up, with error handling for when something upstream goes wrong. If your last leadership meeting included a debate over whose number was right, that's the problem worth solving. devinstance.net/services/autom…

Underrated in .NET 11 Preview 5: the SDK can now run vulnerability and end-of-life package checks at build time. Catch risky dependencies when you build, not when you get audited. devblogs.microsoft.com/dotnet/dotnet-…

Beyond the web, .NET 11 Preview 5 is where C# 15 gets interesting: Union declarations + union patterns Closed class hierarchies LINQ full outer joins JSON Lines (NDJSON) in System.Text.Json Discriminated unions in C#, finally.

.NET 11 Preview 5 is out. The headline for Blazor devs: SSR forms now validate in the browser instantly, no server round-trip. On by default for any SSR form with DataAnnotationsValidator. SSR keeps getting cheaper to use well. devblogs.microsoft.com/dotnet/dotnet-…

.NET 11 Preview 5 is now available! by @dotnet devblogs.microsoft.com/dotnet/dotnet-… #aspnetcore

The runtime story gets the headlines, but .NET 10 and 11 are quietly stacking real #Blazor wins. .NET 10: WASM preloading, passkeys, OpenAPI 3.1 + YAML, QuickGrid RowClass. .NET 11: background services in the browser (IHostedService), QuickGrid row clicks, EnvironmentBoundary. devblogs.microsoft.com/dotnet/dotnet-…

.NET 11 quietly kicked off something big: moving #Blazor WebAssembly off Mono and onto CoreCLR. A WASM-targeting RyuJIT landed in Preview 1. Not production-ready yet, full support is the .NET 12 goal. But the direction is set. Faster execution. One runtime from server to browser. devblogs.microsoft.com/dotnet/dotnet-…

AI made building fast. That's the trap few people talk about. Slow builds let you catch design mistakes gradually. Fast builds just let you ship the wrong idea quicker and burn the budget on it. A short loop without a steering wheel gets you lost sooner. devinstance.net/blog/choosing-…

The product looks polished. The screens render. Stripe takes payments. And the data model underneath might be wrong in ways you cannot see. That gap is the whole problem with a vibe-coded MVP. "It's just an MVP" assumes you know what's underneath. When an AI you didn't supervise wrote it, you don't. You can launch a rough product. You can't launch on a foundation you don't understand. Two hours of a senior engineer's time will surface things you'd otherwise meet under load, in production, with real customers watching. After nearly thirty years cleaning up systems that were fine until they weren't, here's the honest version of what nobody tells founders shipping with Replit, Lovable, or Cursor. Read it here: devinstance.net/blog/vibecodin…

Happy Friday !!!

A new kind of project started showing up in our inbox this year: 'We built this app with AI and it mostly works, but...' Mostly works. That's the phrase every time. The founder vibe-coded an MVP in a weekend. Customers signed up. Then reality arrived: data gets corrupted under concurrent users, there's no real authentication, the codebase is 40 files of duplicated logic no one can follow, and every fix the AI suggests breaks two other things. Here's the part most people get wrong: the answer is usually not 'throw it away and hire devs to rebuild it properly.' The app validated the idea. Real users depend on it. That's worth preserving. What it needs is a rescue, not a funeral. Stabilize the data layer. Add real auth. Untangle the duplication one module at a time while the app keeps running. The same incremental playbook we use on 15-year-old ASP.NET systems works on 15-week-old AI-generated ones. Technical debt is technical debt. If you vibe-coded something that's now carrying real customers and real risk, let's talk before it breaks at the worst possible moment. devinstance.net/services/moder…