Store It. Search It. Analyze It. One Postgres
Your AI app generates three kinds of data at once: events, vector embeddings, and the analytics you run on top of them. The default move is three systems, three sets of tooling, and pipelines to keep them in sync.
TimescaleDB handles all three in the Postgres you already run.
Hypertables ingest events at scale. pgvector and pgvectorscale store and search embeddings. Continuous aggregates keep analytics live without re-querying everything.
Same SQL, same tools, one system to operate. No pipelines to sync, no drift, no second database to feed.
Point it at your heaviest workload and watch it stay fast as data grows. It's still Postgres. Start on Tiger Cloud and get $1000 in credits.
Hi everyone,
This week’s reads focus on the decisions behind reliable .NET applications: how to publish outbox messages, choose a cache, and make uploaded files available across servers.
I’d start with Milan Jovanović’s comparison of Postgres outbox workers and logical replication, then Giorgi Dalakishvili’s DuckDB.NET performance breakdown. Both explain the implementation behind the benchmark results. Anton Martyniuk and Nikola Knezevic add practical guides to caching and file storage.
There’s also a short look at NuGet-delivered agent skills, tomorrow’s .NET Conf, and the support deadline for .NET 8 and 9.
Milan Jovanović
Milan compares concurrent polling workers,
LISTEN/NOTIFY, and reading the Postgres write-ahead log through logical replication. His .NET example publishes outbox messages to RabbitMQ and measures latency, throughput, and what happens after a crash.The WAL reader reduces latency in his tests, but introduces replication-slot management and produces more duplicate deliveries after a hard kill. Both approaches need idempotent consumers. I like that the article examines recovery alongside speed; the WAL implementation is still a prototype, with failover untested.
Giorgi Dalakishvili
Giorgi explains the changes behind DuckDB.NET 1.5.6.1’s improvements to date handling, nullable values, enums and strings. Both versions in his comparison use the same DuckDB engine, which helps isolate the .NET implementation’s effectyou're caching, how many instances run the app, and how much staleness the feature can tolerate., and delete operations using the AWS SDK, strongly typed options,.
In one million-row timestamp benchmark, allocations fall from 2.8 GB to 93 KB. The explanations of boxing, generic virtual calls and date conversion make this worth reading even if DuckDB isn’t in your stack. The results belong to the measured workloads; the techniques give you places to investigate in your own code.
Anton Martyniuk
The second part of Anton’s caching guide focuses on FusionCache and ASP.NET Core Output Cache, then works through choosing a cache for your application.
He covers backplane invalidation across instances, serving stale values when dependencies fail, response-caching policies and tag-based eviction. The decision comes back to practical questions: what are you caching, how many instances run the app, and how much staleness can the feature tolerate? Start there before comparing package features.
Nikola Knezevic
Nikola walks through upload, download and delete operations using the AWS SDK, strongly typed options and Minimal APIs. A small storage service keeps S3-specific requests out of the endpoints, while object metadata preserves the original filename for downloads.
This is a useful starting point when files need to be available across application instances. The tutorial stays focused on the basic integration, with presigned URLs and multipart uploads left for later coverage.
David Grace
David has Claude Code build two skills around an existing ASP.NET Core application: adding an EF Core entity and adding a Minimal API endpoint. Each skill defines the inputs it needs, the project conventions to follow and how to verify the result.
The useful part is the specificity. Table names, column types and string lengths are required inputs, while existing code supplies the patterns. It’s a practical example of turning repeated instructions into a workflow your team can review.
Sukhpinder Singh
A repeated
string.Formatcall can spend time parsing the same template on every iteration. This example moves that work into a reusableCompositeFormatand compares formatting time and allocations over five million calls.It’s a useful technique when a template comes from a variable or resource and gets reused heavily. Start with a measured hot path in your own application; an occasional formatting call is unlikely to justify the extra setup.
James McCaffrey · Visual Studio Magazine
James demonstrates Nadaraya–Watson kernel smoothing, which estimates values using a weighted average of observations. His C# example uses a radial basis function to give more weight to similar inputs, with a bandwidth parameter controlling the smoothing.
A useful read if your application processes measurements or other noisy numerical data. The magazine article appeared November 3; an earlier walkthrough is also available on James’s blog.
Rachel Appel · JetBrains
Rachel’s monthly roundup opens with F# units of measure, then collects community work across .NET, databases, architecture and AI tooling. The links include NativeAOT dependency trimming, primary constructors and agent workflows.
Keep this one for a longer reading session. It’s a broader monthly collection, so some articles predate this week or overlap with earlier editions of The .NET Insider.
Richard Lander · Microsoft
A library’s newest APIs can arrive long before a model knows about them. Richard explores shipping
SKILL.mdfiles alongside the library in its NuGet package, so guidance can match the version your project actually uses.He demonstrates discovery with
dotnet-inspectand copying skills into a repository withdotnet-package-skills. The useful question for maintainers is whether those instructions measurably improve results or reduce wasted work. Bundling a skill alone doesn’t guarantee either..NET Team · Microsoft
The free virtual conference runs November 10–13, starting tomorrow at 16:00 UTC / 8 AM PST. The first two days feature Microsoft product teams and the .NET 11 launch; the final two are community-led.
I’d start with the sessions on C#, ASP.NET Core and Blazor, Aspire, and the .NET AI stack. Wednesday also includes a session on the MCP C# SDK. The agenda lets you select your own time zone.
Microsoft
Both versions reach the end of support tomorrow. Applications will continue to run, but regular support and security servicing for those releases end.
If either is still in your production inventory, check the migration plan across SDKs, dependencies, build pipelines and hosting environments. For teams choosing an LTS target, .NET 10 remains supported through November 14, 2028.
Which of these approaches are you using in your own applications? I’d like to hear what has worked well and what you would handle differently next time.
Thanks for reading,
Get future editions at dotnet.news.
Have an article to suggest? Send it to [email protected].



