Jim Togher retweeted
Daylight reveals the extent of damage caused to Launch Complex 36 (LC-36) and the surrounding area of Cape Canaveral Space Force Station in Florida, following last night’s massive explosion of Blue Origin’s New Glenn during a Static Fire Test. Significant fire damage to the launch pad, tower, and other infrastructure can be seen - which will undoubtably require months of repairs - while debris from New Glenn lay scattered around LC-36.
Photo credit: @asherbphotos @tweetsiphotos @LaunchHeavenX
Inventory that's accurate to the second. POS that doesn't blink during peak. Mobile apps that talk to fulfillment that talks to the store.
Engineers from @Walmart, @TheHomeDepot, @Starbucks, and @AceHardware are joining us at RethinkConn 2026 to talk about how it actually gets built — at scale, across thousands of locations, in real time.
If your team runs anything connected, this is the room.
Half-day, virtual. 👉 synadia.com/lp/rethinkconn-2…
#RethinkConn2026 #RetailTech #ConnectedRetail
We're proud to share that Bruno Baloi has been recognized as an "AI Journal Thought Leader" — and his latest piece is one you'll want to read.
In "Living on the Edge: Rethinking Event-Driven Architecture for Distributed Intelligence," Bruno explores how the shift toward distributed computing is fundamentally changing the way systems must communicate, recover from failure, and make decisions in real time.
Some key takeaways:
🚀 Edge AI isn't just about speed
⛔️ In distributed environments, failure isn't the exception
📏 Event-driven systems are a natural fit for the edge, enabling resilience through asynchronous messaging and decoupled design
🐦🔥 The line between "cloud" and "edge" matters less than the ability to design systems that operate seamlessly across both.
At Synadia, this is what we think about every day — building the connective tissue for distributed, intelligent systems that work anywhere, reliably.
Read Bruno's full article here 👇
aijourn.com/living-on-the-ed…
#DistributedSystems #EdgeComputing #EventDrivenArchitecture #NATS #AI #Synadia
A platform bridging edge and core should support four flow control capabilities:
1. Subject/topic filtering
2. Subject mapping / traffic shaping
3. Dynamic load distribution (weighted routing)
4. Varied consumption models (push vs pull)
synadia.com/resources/living…
NATS has a second monitoring model most people don't use.
The system account publishes events — connections, disconnections, auth failures, JetStream operations — the moment they happen.
Push-based, not pull-based. No scrape interval, no blind spots.
Salute to Robert Duvall. GF, Appoclps Now, Falling Down, COLORS, The Apostle, Days of Thunder, The Great Santini, Assassination Tango, Gone In Sixty Seconds, TRUE GRIT. My gosh the list goes on. A true iconic American actor. His kind will be missed.
2
67
A little less sexual and just a little bit of Jesus, and it would have been cultural and perfect.
1
13
Jim Togher retweeted
One of the many "unknown" super-powers..
"Where did my message go?"
In distributed systems, poor visibility turns small issues into long incidents—silently burning engineering time and delivery velocity.
NATS solves this by making message tracing a native capability. Since 2.11, a single command reveals how a message traverses clusters, gateways, and leaf nodes—end to end.
No external observability stack.
No operational add-ons.
No instrumentation burden on application teams.
With:
nats trace demo.subject "hello"
operators and devs can see exactly where messages went without adding complexity to the stack.
Stop chasing messages. Start shipping.
AI systems that interact with the physical world have a blind spot:
the boundary between cloud and edge.
Getting data from a sensor to a model is one problem. Getting a decision from a model back to an actuator is another.
Most architectures solve one direction well and treat the other as an afterthought.
But AI systems that interact with the physical world don’t get that luxury.
They have to close a control loop.
Events up:
Telemetry from thousands of sensors, distilled into meaningful signals at the source, flowing securely to wherever inference happens.
Commands down:
Decisions translated into actions and pushed back to the edge—reliably, securely, and fast enough to keep up with physics.
A turbine doesn’t care how good your model is if the throttle command arrives 500 ms late.
A robot arm doesn’t care about your 99.9% uptime if the 0.1% happens mid-motion.
This is where many architectures fail.
Stacks that stitch together Kafka for ingest, MQTT for delivery, Redis for state, and a custom edge gateway often look fine on arch diagrams.
But they break down when used as a true control plane.
Every seam adds latency.
Every translation adds failure modes.
Every boundary expands the security surface.
The cloud–edge handoff isn’t an implementation detail you can “solve later.”
For physical-world AI, it is the architecture.
The teams winning at edge AI aren’t the ones with the best models.
They’re the ones that built a real nervous system:
one architecture - bidirectional - secure - and fast enough to act while it still matters.
That's what we're building at Synadia with @nats_io
Jim Togher retweeted
"Nothin’ but NATS” used to be just a tagline. Now it’s an architecture strategy.
As 2025 winds down, we’ve been reflecting on why @nats_io is different—and what that unlocks.
3 core principles tell much of the story:
1. Location independence
Services don't need to know where other services live.
No load balancers.
No DNS gymnastics.
No service mesh ceremony.
Just connect and communicate. Add, remove, migrate, or redeploy services without reconfiguring clients or rewiring infrastructure.
2. Many-to-many by default
Most systems optimize for 1:1 communication and treat fan-out as an exception. Every new consumer adds more plumbing.
What if M:N was the default?
When one signal can naturally reach many consumers, systems gain flexibility.
New capabilities show up as subscribers. Old ones disappear without coordination.
Features evolve independently, scale becomes additive, and change stops flowing through a single critical path.
3. Async-first
Temporal decoupling isn’t a feature. It’s the bedrock of system resiliency.
Async foundations absorb spikes, isolate failures, and prevent cascades—while still supporting request-reply when you need it.
These aren’t new ideas. They’re the fundamentals NATS was built on.
Teams usually adopt NATS for one use case.
Then it replaces their message queue.
Then parts of their cache and coordination layer.
Then their service discovery.
“Nothin’ but NATS” used to be just a tagline.
Now it’s an architecture strategy.
Sometimes the best move isn’t adding more—it’s getting back to the basics.
Jim Togher retweeted
Down and out. Normally, not a good thing.
But in distributed systems, down and out is where the world is headed.
Down to smaller devices. Out to environments beyond data center walls.
Is your streaming and messaging infra ready for that trip?
---
If you're all-in on Kafka, the honest answer is "probably not".
Kafka is an elephant. A behemoth. It was built for the data center and it's not leaving. The architecture won't let it.
Kafka excels in centralized environments—but its assumptions don’t always extend to the edge.
The JVM alone carries a meaningful baseline memory cost, which is fine in the data center but impractical on compute-constrained devices deployed outside it.
Kafka also assumes brokers are continuously reachable and tightly coordinated. In environments with intermittent connectivity—remote sites, mobile systems, unreliable WANs—that model becomes too fragile.
And while the operational overhead is manageable for a few clusters, it compounds quickly when you try to run many independent deployments (e.g. at the edge).
Kafka was built for scale up—bigger clusters, more throughput, more partitions.
It was never designed for scale down or scale out beyond data center walls.
---
@nats_io was.
A single, sub-20MB binary. No JVM. No external dependencies.
Runs the same whether it's in AWS, in your data center, on a factory floor, or embedded in a device.
Leafnodes let edge locations operate independently—process locally, decide locally, sync when connected.
No partition rebalancing. No ZooKeeper migrations. No dedicated specialist required.
One platform for messaging, queuing, steaming, KV and more. One operational model. From cloud to your user-defined edge.
---
The data center era built cathedrals. Massive, impressive, rooted in place.
(And they'll mostly stick around—NATS at the edge can feed your kafka→flink→data lake pipeline.)
But in the worlds of edge and AI, you need something that travels light.
So remember: being down and out doesn't mean you're struggling!
It means your architecture is ready for the edge computing future.
Jim Togher retweeted
It has been an amazing partnership with the @Rivian team pushing innovation into connected cars. Many of @synadia customers term us the "decentralized nervous system" - in this case that is very true.
linkedin.com/posts/waldemarq…
Jim Togher retweeted
Tomorrow @nats_io turns 15!
We're celebrating with gifts... for you! Here's how it works:
1. Post here on X about why NATS is a great fit for your specific use case. The more details the better!
2. Tag the @nats_io and @synadia accounts in your post
3. We'll message you with a link to choose your NATS hat or shirt!
---
The fine print:
- some geographies may not be supported by our vendor
- quantities are limited, but we'll do our best to reward all good faith posts

















