Samuel Tambunan's Blog

Weekly .NET & AI Briefing — September 7–13, 2026

A lighter week for the big consumer engineering blogs (Netflix and Airbnb haven’t posted anything new since late August), but a strong one for .NET itself and for the “how do we actually build software with AI agents” conversation that keeps following David Fowler, Martin Fowler, and Jimmy Bogard around. .NET 11 hit Release Candidate 1, C# 15’s union types are getting real documentation, and Uber published three meaty backend deep-dives in a single week. Thirteen links below, roughly in order of how directly they’ll help you toward the architect/leader track.

.NET & C#

  1. Announcing .NET 11 Release Candidate 1 — .NET Blog, September 8 .NET 11 RC1 ships with a go-live license, so it’s supported in production ahead of GA. Highlights include JSON and TLS session improvements, FP16 hardware support in the runtime, better crash reporting, and C# 15 stabilizing union types. Worth a skim just to see what’s landing in the next LTS-adjacent release.

  2. Use C# unions and closed hierarchies in ASP.NET Core — .NET Blog, September 10 Dmitrii Korolev walks through how C# 15’s new union types and closed class hierarchies serialize with System.Text.Json across Minimal APIs, MVC, SignalR, Blazor, and OpenAPI. If you’ve been modeling “one of several alternatives” with inheritance hacks or discriminated-union libraries, this is the shape of the built-in answer.

  3. .NET and .NET Framework September 2026 servicing releases — .NET Blog, September 8 The routine but necessary one: patch notes for .NET 8, 9, 10, and .NET Framework this month. Five minutes to confirm nothing in your current LTS needs urgent attention.

  4. The pain of serializing unions and closed class hierarchies with System.Text.Json — Andrew Lock, September 1 The same C# 15 union feature from the trenches — Andrew Lock’s usual deep, slightly skeptical dig into the rough edges before you hit them yourself. Good pairing with link #2 above if you want the “here’s what the docs don’t tell you” version.

  5. Let Postgres Enforce Tenant Isolation — Milan Jovanović, September 12 A pattern every architect building multi-tenant systems should have in their back pocket: use PostgreSQL row-level security as a database-enforced backstop alongside EF Core global query filters, so a bypassed or forgotten filter still can’t leak another tenant’s data.

Architecture & Leadership

  1. Vertical Slice Architecture Webinar Recording, and What’s Next — Jimmy Bogard, September 1 Jimmy Bogard’s recap of a 700-person webinar on Vertical Slice Architecture, with a line worth sitting with: “structure matters more now, not less” — because AI makes generating code cheap but verifying it stays expensive. He’s following up with live sessions and an in-person workshop if you want to go deeper.

  2. Fragments: September 8 — Martin Fowler, September 8 Martin Fowler’s weekly link roundup this time centers on the gap between AI systems generating output and anyone actually verifying it — what he calls “counterfeit utility” — plus reader fatigue with obviously LLM-written content. Short, but the kind of thing worth quoting in a team discussion about AI-assisted delivery.

  3. Microsoft’s Fowler Says Typing Code Is Over — implicator.ai, September 6 Coverage of David Fowler (.NET Aspire) arguing that manually typing code is a shrinking part of the job, with Aspire’s own workflow — agents compiling, running, and testing while humans review and decide — as the model. Aspire PM pushes back a little too: “generating code and shipping working full-stack apps are very different things.”

  4. .NET Conf 2026 Community Days — Call for Presenters — .NET Blog, September 8 Submissions for community-led .NET Conf sessions are open through October 6. If part of “leader” is building a public profile, a Community Days talk is a low-friction way to start.

Engineering at Scale

  1. Halving the Time: How Uber Eats Rebuilt Its Search Pipeline — Uber Engineering, September 10 Uber Eats cut search latency by 50% by cutting dead retrieval paths, splitting hydration into parallel phases, and — notably — using an AI agent workflow to systematically hunt down bottlenecks. A good real-world example of AI-assisted performance work, not just AI-assisted code generation.

  2. From One Horizontal Scaling Controller to Many: Evolving Uber’s Compute Platform — Uber Engineering, September 9 Uber’s redesign of its Kubernetes autoscaling control plane to let multiple orchestrators safely scale workloads at once, including hard-won lessons on stale caches and multi-writer consistency. Solid case study material if you’re ever arguing for stronger consistency guardrails in a distributed system.

  3. From Signals to Context: Lessons from Lavaredo Ultra Trail — Uber Engineering, September 11 Uber’s third post of the week, on turning raw event signals into usable context — worth a read if you’re working on anything observability- or context-pipeline-shaped.

Bonus

  1. .NET AI Essentials — The Core Building Blocks Explained — .NET Blog Not brand-new this week, but a good standing reference if you’re still getting oriented on Microsoft.Extensions.AI and how Microsoft wants .NET developers building AI features. Worth bookmarking alongside this week’s links.

A quieter week from Netflix and Airbnb — their engineering blogs’ most recent posts are from late August, so nothing new there to report; check back next week.