Microsoft has rolled out new storage improvements in the latest Windows 11 Insider builds, including Dev build 26300.8170 and Beta build 26220.8165. These changes focus on faster storage settings, a better user experience, and removing an artificial limitation that existed for decades.

The update improves performance when navigating Settings > System > Storage > Advanced Storage Settings > Disks & Volumes, which is where you check drive properties, partitions, and storage details. Microsoft has also changed how the Storage page handles permissions, so the UAC prompt no longer appears immediately and only shows up when accessing temporary files.

Alongside that, Windows now allows formatting FAT32 drives up to 2TB via the command line, replacing the long-standing 32GB limit that existed for decades despite the filesystem supporting much larger volumes internally.

Opening disk properties on large drives has been unusually slow for a long time, especially on systems with multiple or high-capacity volumes.

I tested this myself on multiple drives, and the difference is bigger than I thought.

Windows 11 finally fixes slow storage settings on large drives

Changing drive properties from the Settings app wasn’t something I used to do, because my muscle memory clicks the Windows key and X, opening the Disk Management tool.

Although it looks decades old, which it is by the way, it still gets the job done. But I understand that I’ll have to force myself to use the latest tools if I want to see Windows lose legacy code.

That said, slow performance when going to see storage on large volumes from the Disk & Volumes page is a major annoyance. If you have a system with larger drives, especially mechanical HDDs with multiple partitions, opening Disks & Volumes and clicking on a drive’s properties could take an unusually long time to load.

In my case, it consistently took around 15 seconds for the page to fully respond.

Here is my PC opening the Storage settings for a 130GB Drive:

Yes, it took almost 15 seconds. Here’s the same thing on a 292GB drive:

Checking a third time with a 409GB drive:

All this was done on a 1TB mechanical hard drive split into multiple partitions, with different types of files across them. My PC itself wasn’t under load, and there weren’t many background apps running.

The issue becomes worse when you have larger volumes or multiple drives connected, because the Settings app has to query more data before rendering the UI.

Fortunately, Microsoft has finally improved the performance of Storage settings.

Microsoft’s release notes for the Dev build 26300.8170 and Beta build 26220.8165 simply mention “improved performance when navigating storage on large volumes,” but the actual change is more significant. The delay when opening disk properties is almost gone in the new Insider build.

To confirm that this wasn’t just hardware differences, I tested the same scenario inside a virtual machine running the latest Insider build. This VM only has 4GB RAM and 2 CPU cores, far less capable than my main system. Yet, opening disk properties was almost instant.

Yes, it’s instant. And I wasn’t expecting a massive improvement like this.

So why was this slow before? The likely cause comes down to how the modern Settings app handles storage data. Unlike legacy tools, it uses newer UI layers and background queries to fetch disk information, partitions, file system details, and usage data.

On larger drives, especially HDDs, this can take longer due to slower read speeds and the amount of metadata involved. If those queries are handled inefficiently or synchronously, the UI ends up waiting instead of rendering immediately.

This seems to be what Microsoft has addressed here. The faster response may be due to better handling of data fetching and UI rendering, possibly by optimizing how disk information is loaded or by reducing blocking operations.

Alongside this performance improvement, Microsoft is also updating a long-standing limitation in Windows file systems.

FAT32 limit jump from 32GB to 2TB is an important change in Windows

Until now, Windows has only allowed formatting FAT32 drives up to 32GB using its built-in tools, which wasn’t because of the file system itself. FAT32 has always supported much larger volumes, theoretically up to 2TB with standard sector sizes. The restriction was something Microsoft imposed inside Windows.

With this update for Windows Insiders, that artificial limit is gone, at least through the command line. You can now format FAT32 volumes up to 2TB natively without relying on third-party tools.

The original 32GB cap goes back to a time when large drives were rare and FAT32 wasn’t considered efficient at scale. It suffers from fragmentation and lacks features like journaling, which makes it less reliable compared to NTFS. Microsoft’s approach back then was to push users toward NTFS for internal drives and later exFAT for removable storage.

Their strategy worked, but it also meant that anyone who needed FAT32 for compatibility had to jump through unnecessary hoops.

Even today, FAT32 is still required in a surprising number of scenarios. Firmware updates for motherboards usually need a FAT32-formatted USB drive. Some gaming consoles and media devices also expect FAT32 for detection. Embedded systems and older hardware still depend on it as well.

That said, FAT32 comes with its own constraints. The most obvious one is the 4GB file size limit, which makes it impractical for modern workloads like large video files or backups. It’s not a replacement for NTFS or exFAT, and it’s not meant to be.

Storage settings now avoid unnecessary UAC prompts

Earlier, opening Settings > System > Storage could immediately trigger a UAC prompt, even if you were just trying to view basic storage information, which didn’t make much sense for a read-only action.

Now, the UAC prompt only appears when you access areas that require elevated permissions, such as viewing or managing temporary files.

Bigger improvements are coming soon to Windows 11

These storage improvements are currently available in Windows 11 Insider builds (Dev 26300.8170 and Beta 26220.8165). Microsoft hasn’t shared a timeline for stable rollout, but changes like these usually make it to regular users in the next few months or maybe even weeks.

With all the hatred from social media, I was, weirdly enough, not expecting Microsoft to be skilled enough to fix a part of Windows 11 that made it feel slow. Now that they have proved me wrong, it gives me hope that the development team can pull off everything that Microsoft is planning to fix in Windows 11.

The post Tested: Windows 11 just fixed slow storage management and removed a 30-year FAT32 limit appeared first on Windows Latest

"Rust's rise shows signs of slowing," argues the CEO of TIOBE. Back in 2020 Rust first entered the top 20 of his "TIOBE Index," which ranks programming language popularity using search engine results. Rust "was widely expected to break into the top 10," he remembers today. But it never happened, and "That was nearly six years ago...." Since then, Rust has steadily improved its ranking, even reaching its highest position ever (#13) at the beginning of this year. However, just three months later, it has dropped back to position #16. This suggests that Rust's adoption rate may be plateauing. One possible explanation is that, despite its ability to produce highly efficient and safe code, Rust remains difficult to learn for non-expert programmers. While specialists in performance-critical domains are willing to invest in mastering the language, broader mainstream adoption appears more challenging. As a result, Rust's growth in popularity seems to be leveling off, and a top 10 position now appears more distant than before. Or, could Rust's sudden drop in the rankings just reflect flaws in TIOBE's ranking system? In January GitHub's senior director for developer advocacy argued AI was pushing developers toward typed languages, since types "catch the exact class of surprises that AI-generated code can sometimes introduce... A 2025 academic study found that a whopping 94% of LLM-generated compilation errors were type-check failures." And last month Forbes even described Rust as "the the safety harness for vibe coding.." A year ago Rust was ranked #18 on TIOBE's index — so it still rose by two positions over the last 12 months, hitting that all-time high in January. Could the rankings just be fluctuating due to anomalous variations in each month's search engine results? Since January Java has fallen to the #4 spot, overtaken by C++ (which moved up one rank to take Java's place in the #3 position). Here's TIOBE's current estimate for the 10 most popularity programming languages: Python C C++ Java C# JavaScript Visual Basic SQL R Delphi/Object Pascal TIOBE estimates that tthe next five most popular programming languages are Scratch, Perl, Fortran, PHP, and Go.

Read more of this story at Slashdot.

Microsoft hasn’t had a coherent GUI strategy since Petzold
8 minutes by Jeffrey Snover

Over decades, shifting frameworks like Win32, WPF, UWP, and WinUI—often disrupted by internal conflicts and changing priorities—have confused developers. Jeffrey says that the problem is not weak technology, but poor coordination and frequent strategic pivots that leave no single, reliable path forward.

Share your thoughts on .NET dev tooling and get a chance to win a reward
sponsored by Jetbrains

JetBrains is conducting a survey to better understand how .NET dev tools impact developer productivity, and your perspective could really help the company make its tooling even better. It takes about 10 minutes to complete and focuses on your real-world experience with the tools. As a thank-you for your time, you’ll have the chance to win a small prize. Take the survey and make your voice count!

C# 14: Introducing interceptors
5 minutes by Anthony Giretti

C# 14 introduces interceptors, a way to silently replace a specific method call with a different one at compile time, with no runtime cost. The compiler swaps the call before producing any binary, making it fully compatible with Native AOT and leaving the developer's source untouched. Interceptors are always generated by Source Generators, never written by hand, since they rely on encoded file positions that change with every edit.

How to delete and update millions of rows in EF Core efficiently
7 minutes by Chris Woodruff

Loading rows into memory before modifying them works fine in development but falls apart at production scale. EF Core 7 added ExecuteUpdate and ExecuteDelete, which run directly on the database without loading any objects, cutting memory use and round-trips dramatically.

Your logout button is lying: ASP.NET session security done right
7 minutes by Martin Stühmer

Martin explains how poor session setup in ASP.NET can create serious security risks. He contrasts common mistakes with secure configurations that meet audit standards. Key practices include setting idle and absolute timeouts, using secure cookie settings, properly invalidating sessions on logout, and rotating JWT tokens. Martin points out that the goal is to move from “it works” to truly secure session management that prevents breaches and passes audits.

Functional programming in C# 15: Union types and the ApiResult Monad
9 minutes by Tore Aurstad

C# has been steadily absorbing ideas from functional programming — pattern matching, records, immutability. With C# 15, we get the feature that ties it all together: union types. Tore walks you through why they matter, how they work, and how they enable a clean ApiResult result monad that eliminates try/catch boilerplate and makes error handling composable.

And the most popular article from the last issue was:

• Getting started with PgVector in .NET for simple vector search

Getting started with PgVector in .NET for simple vector search
8 minutes by Milan Jovanović

Not every AI feature needs a dedicated vector database. pgvector is a PostgreSQL extension that adds vector storage and similarity search directly to your existing database. You enable the extension, create a vector column, and start querying. In this post Milan shows you what vector search is and when you need it, he walks you through provisioning pgvector with .NET Aspire and Ollama, generating embeddings with MEAI and storing with Dapper and querying by semantic similarity using cosine distance.

There Are Monsters Inside This Low-Level Book About Processors
sponsored by Kevin Focke

TUnit now captures OpenTelemetry traces in test reports
5 minutes by Tom Longhurst

TUnit now captures OpenTelemetry traces directly inside its HTML test reports. When a test runs against an ASP.NET Core app, spans from database queries, HTTP calls, and middleware all appear as a timeline next to the test result. Setup takes one NuGet package and two base classes. Tom says this makes failures easier to diagnose since instead of guessing, you can see exactly what the app did during the failing test.

Generic EqualityComparer for classes in C#
12 minutes by Tore Aurstad

C# classes use reference equality by default, meaning two objects with the same data are not considered equal. GenericEqualityComparer fixes this by comparing class instances by their property and field values instead. It uses reflection at startup to build fast compiled delegates, and supports public and private members. Tore suggests to avoid it in performance critical loops, and he prefers built-in equality for records and structs.

How to seed data to EF Core
6 minutes by Ricardo Peres

There are four main ways to seed data in EF Core. You can insert data explicitly in code, write a data-only migration with raw SQL, declare seed data inside entity configuration using HasData, or register seeding callbacks with UseSeeding and UseAsyncSeeding. Ricardo points out that each approach suits different needs: HasData works best for small fixed datasets, migrations for one-time SQL inserts, and UseSeeding for flexible logic that runs with both migrations and EnsureCreated.

URL pattern matching in .NET
2 minutes by Gérald Barré

Working with URLs in .NET applications often involves pattern matching for routing, security policies, or content filtering. While you could write custom regex patterns or string comparisons, Gérald argues there's a better way: the Meziantou.Framework.Uri package brings standardized URL pattern matching to .NET. This package implements the WHATWG URL Pattern API specification, providing a consistent and powerful way to match URLs against patterns with parameters, wildcards, and modifiers.

And the most popular article from the last issue was:

• What your .NET exceptions are telling attackers (and how to stop it)

Building a VS Code-style extension system in C# means implementing extension points, activation events, and contribution points. Complete .NET 8 guide.

Thanks to a tip from Rahim Nathwani, here's a uv run recipe for transcribing an audio file on macOS using the 10.28 GB Gemma 4 E2B model with MLX and mlx-vlm:

uv run --python 3.13 --with mlx_vlm --with torchvision --with gradio \
  mlx_vlm.generate \
  --model google/gemma-4-e2b-it \
  --audio file.wav \
  --prompt "Transcribe this audio" \
  --max-tokens 500 \
  --temperature 1.0

I tried it on this 14 second .wav file and it output the following:

This front here is a quick voice memo. I want to try it out with MLX VLM. Just going to see if it can be transcribed by Gemma and how that works.

(That was supposed to be "This right here..." and "... how well that works" but I can hear why it misinterpreted that as "front" and "how that works".)

Tags: uv, mlx, ai, gemma, llms, speech-to-text, python, generative-ai