"How Do You Fix 70% Data Loss Across 1 Million Concurrent Connections?" - this is a question a Production @AkkaDotNET Support customer asked us earlier this year. We solved it.

Here's how we did it: petabridge.com/blog/why-buy-a…

In about ~2 hours I'll be running our live "Introduction to @AkkaDotNET Streams" training and this session is open to the public. Anyone can join without a support plan.

It's listed at the top of petabridge.com/training/?utm_…

Going live with our November 2026 @AkkaDotNET Community Standup in about 15 minutes - lots of important v1.6 work plus Akka.Reminders!
youtube.com/watch?v=SqvtWg…

Your actors sleep. Their reminders don't.

Akka.Reminders gives @AkkaDotNET sharded actors durable, re-entrant scheduling — the reminder wakes the entity even after passivation, a node failure, or a full process restart.

Live tomorrow, 12 PM CT 👇
youtube.com/watch?v=SqvtWg…

Working on Quic support for @AkkaDotNET v1.6

Our new TCP transport already supports logical multiplexing using Akka.Streams, which solves head-of-line blocking problems in a few key ways (big messages get chunked)

This will do multi-plexing at the transport layer itself!

Lots very inefficient compared to the TCP transport, but we're just at the starting line now

If you're polling a status endpoint, you probably want a stream.

That's the whole decision rule. Polling is a confession that your return type threw away the information you needed.

55 minutes, three production systems, and the honest limits of the construct:
youtu.be/2fqZsF9y4l8

Subscribe for more .NET architecture content.

Task.WhenAll is all or nothing, at the end.

Three fulfillment operations run. One fails. You get an AggregateException and no idea which of the other two finished.

Streams turn that into a decision: two of three entitlements are live, the third opened a support ticket automatically, and the customer can see all of it.

Same events also give you an audit trail, a support view, and a debugging log. One stream, three stories.

youtu.be/2fqZsF9y4l8

Task<T> has three states: pending, completed, faulted.

"Alive and making progress" and "stuck forever and never coming back" are the same one.

That's not an observability gap you can instrument away. It's how Task is designed.

New on the channel, 55 minutes on what to do about it:
youtu.be/2fqZsF9y4l8

Subscribe if you want more like this.

Just posted my talk from last Friday, "You Should Be Using IAsyncEnumerable," on @petabridge 's YouTube channel: youtube.com/watch?v=2fqZsF…

Give it a watch and leave a comment. Let me know what you liked and what we could do better too

Kafka scales a consumer group by partition. If you want partitions processed in parallel while ordering stays intact within each one, Akka.Streams.Kafka's CommittablePartitionedSource is built for that shape.

It also had a throughput defect. Fixed in 1.5.73: ~6,200 → ~16,900 8m8UUuvk

Tighter by default. Netclaw 0.28.0 tightened security across the board:

• Agent tools can't reach ~/.ssh, ~/.aws, ~/.kube, ~/.netrc, or ~/.docker/config.json
• The daemon checks Host and Origin on loopback requests
• Tool access is scoped to each turn's audience
• The output ecrets

netclaw.dev/changelog/0.28…

As I mentioned in our webinar today, we've opened our Nov 12th @AkkaDotNET streams training course (2 hours) to the public - free for anyone to attend.

You can register here: attendee.gotowebinar.com/register/62270…

Netclaw's approval system is even better.

• Recognizes more previously-approved command shapes
• Better sensitive data filtering

netclaw.dev/changelog/0.28…

Join us here: youtube.com/watch?v=bKZzju…

Tomorrow at 12pm Chicago time: October 26' @AkkaDotNET Community Standup featuring our v1.6 betas:

- AOT support
- Source-generated serialization
- Radical performance improvements to Akka.Remote

Join us live or later - I'll be doing a developer walkthrough and taking questions!

Ask it for harder work. Netclaw 0.28.0 runs longer sessions with no tool-call limits, takes your input mid-run, and delegates to background sub-agents — so you can hand it more complicated jobs.

netclaw.dev/changelog/0.28…

Please join us on Friday, October 30th 11:00am Chicago Time for "You Should Probably Be Using IAsyncEnumerable"

Web programming paradigms increasingly support and are moving towards streaming application models, which solve many problems that legacy and default request-response , such as polling.

Come spend 60 minutes with us to learn about how IAsyncEnumerable and streaming architectures can help your projects run at lower latencies, use less resources, support richer interaction patterns, and more.

petabridge.com/webinar/stream…

Friday, October 30th: I'll be hosting a quick webinar entitled "You Should Probably Be Using IAsyncEnumerable"

While we do talk about IAsyncEnumerable, of course, this is really more about moving away from request / response architectures everywhere to using streaming.

1/3

So @AkkaDotNET v1.6 is already an order-of-magnitude improvement for networking and serialization performance over v1.5, but the next beta of it is about to be even faster:

Looking at about a ~20% throughput improvement across the board and significant reductions in CPU / mem All from simple mechanical fixes plus some experimental IActorRef compression on the receiving side of the wire.

An LLM's bug report is an opinion. A model-based test that runs 20,000 random sequences is one it can't easily dodge.

That's how we checked Claude's 5 findings on Akka.Reminders. Prerequisite: a written spec.

youtube.com/watch?v=YOqLRH…

Claude reported 5 consistency bugs in Akka.Reminders, one of our Akka.NET plugins.

2 were real spec violations. The other 3 were gaps where our own spec was silent.

How we verified them: model-based testing with CsCheck.

youtube.com/watch?v=YOqLRH…