Jake Page retweeted
If your services run on @awscloud ECS, you know the loop. You build the image, push it, register a new task definition and wait for the rollout. Only then do you find out whether your change works, and if it doesn't, you do it all again.
This week we launched mirrord for ECS which collapses the loop. Write code, then immediately test in staging.
@EBukchin wrote about how it works and what testing on ECS looks like now: metalbear.com/blog/stop-deplβ¦
#aws #ecs #containers
Jake Page retweeted
Coding agents write a change in minutes. Testing it on AWS ECS still takes hours.
Every team on ECS we've talked to describes the same loop: build the image, push it, register a new task definition revision, wait for the rollout, check what happened, and do it all again if it's wrong.
That was already too slow for humans, but it's so much worse for agents who write a change in seconds, then have to wait a full rollout to learn whether it worked.
Most of the speed you got from the agent ends up spent on waiting for rollouts.
On Kubernetes, mirrord fixed this a while ago. 20,000+ developers use it along with their AI agents to test generated code against their real staging environment without deploying it.
Today we're going beyond Kubernetes with the launch of mirrord for ECS.
β’ Your code runs on your laptop or in an agent's sandbox, but it gets real traffic, environment variables, network access, and files from your ECS task in staging. You see if a change works in seconds, not after a rollout.
β’ To set it up, it's just one task definition change.
β’ Only requests that carry your session header come to your code, so you, your teammates and your agents can each test your own version of the service at the same time, on the staging environment you already share.
mirrord for ECS is live today, and itβs free to get started with! Let your agents test against real staging without deploying, and without getting in anyone else's way.
See you in SLC π»
If mirrord worked on people, we'd stay home and still get all of @KubeCon_'s and @TheLeadDev's traffic. It doesn't, so we're getting on planes.
Next week we'll be at @CloudNativeFdn's KubeCon North America in Salt Lake City (booth 352, Nov 9β12) and LeadDev Berlin (booth 103, Nov 9β10). Come talk to us about giving your AI agents a real cluster to test on, or just grab a mirrord eye mask before they're all gone!
Going to LeadDev Berlin? Use code METALBEAR for β¬200 off your ticket: leaddev.com/leaddev-berlin/?β¦
What's mirrord?
mirrord is a Kubernetes development platform that connects AI agents and developers to a live cluster so they get real-world context and a real feedback loop. Your service runs wherever you're working, locally, in CI, or in an agent's sandbox, while mirrord proxies its traffic, environment variables, and files to and from a shared staging environment, so it behaves as if it were deployed without actually being deployed.
1
41
Jake Page retweeted
Happening tomorrow!
Uzi Abdullayev (Senior Software Engineer) and @TomerBrook (Senior AI Engineering Manager) from @mondaydotcom are joining @EBukchin (CTO, MetalBear) to talk about how @mondaydotcom scaled agentic development with an in-house built platform!
They'll be talking about the details of:
πΉ How agents pick up tasks from boards and take them all the way to a tested PR
πΉ The feedback loop that makes those PRs worth merging, including testing against real services with mirrord
πΉ What they learned scaling this approach to thousands of agents
Sign up here: metalbear.com/events/scalingβ¦
Jake Page retweeted
World class engineering teams like @PodiumHQ βs use mirrord (by @metalbear ) to power their autonomous coding agents.
If youβre serious about building an incomparably fast agentic workflow that delivers high quality software with minimum human intervention, you should too.
We built a $150M ARR business twice.
The first time took 8 years. The second took 2.5.
Today, Podium crossed $150M in AI annual recurring revenue.
Three years ago, we already had a great software business. Well over $100M in revenue, tens of thousands of customers, and a product we'd spent nearly a decade building.
And we decided to start over.
The business was performing well. But we believed AI was going to fundamentally change the software industry and make much of traditional software obsolete.
As a founder, you pour your life into what you're building. An insane amount of time, sacrifice, and effort. The hard part is not getting too attached to it.
When we started building our first AI agents, we had to confront the possibility that much of what we'd built might not be that valuable in the future. That's a tough pill to swallow.
But we also realized those same 10 years had put us in a unique position to take advantage of what AI could do for our customers.
Instead of protecting what we'd spent a decade building, we decided to bet the future of the company on AI employees that actually do the work for our customers.
We massively underestimated what we'd signed up for.
We ended up building entirely new operating systems for the industries we serve, from the AI agents all the way down to the systems of record. Now customers run their entire businesses on Podium.
Reinventing our product wasn't even the hardest part. We had to reinvent how the entire company operated.
When you've been running a company for over a decade, the operational inertia is incredibly powerful. We had to rethink how we build products, how we sell, how we serve customers, and how we run the business.
We effectively started a second startup inside an already successful company. And we're still at the very beginning.
Today, we have 20,000 AI agents working for our customers. They've helped generate more than $13 billion in revenue for those businesses. And we're barely scratching the surface of what AI will be able to do for them.
And we've done all of this while becoming profitable and generating free cash flow. We're building Podium to be around for decades.
Huge credit to our team for having the guts to make this bet and the ability to actually pull it off. We still have a long way to go and tons of problems to solve, but it's so damn fun to be winning when the opportunity ahead is this big.
In the age of AI, you can either disrupt your own company or wait for somebody else to do it.
I'd rather do it ourselves.
Jake Page retweeted
We wrote about how our engineering team uses AI back in January. Nine months is a long time in this space, so @EBukchin, our CTO, revisited it with what's actually changed since.
Read here to find out which tools we're using now, and where the struggles have shifted to: metalbear.com/blog/engineeriβ¦
Jake Page retweeted
Replying to @deedydas
@deedydas suggest solving it with old ways - more compute, more efficiency, more cache. Beyond amount of CI needed, there's almost no value in UT when using AI while tremendous value lies in E2E.
None of the listed companies help you run E2E better/faster.
The biggest cost of E2E is spinning up a real-environment.
our (@metalbear) approach with mirrord offers something different - give each test/ai/engineer a slice of a real environment, isolating flows allowing infinite scale while almost zero cost (beyond compute needed for the tests). Use it in CI, in CD and in the cloud agents it selves to have everyone iterating against same baseline.
for example - metalbear.com/mirrord/case-sβ¦
or @fagamericano who writes a lot about it (also mirrord user :D)
As cost of coding has plummeted, long builds (CI) has become a bottleneck for many software cos. Three of the biggest startups solving this:
1. Blacksmith (@useblacksmith)
2. Namespace (@namespacelabs)
3. Depot (@depotdev)
Jake Page retweeted
Take a read π
This week @dorco8 and @EBukchin's webinar left such an impression on me that I ended up writing a blog about :P
If you missed the webinar (or prefer text over video) and want to learn how @mondaydotcom runs its agent evals read the blog here: metalbear.com/blog/monday-agβ¦
Jake Page retweeted
"I feel more comfortable in what I ship now because we can test our code in staging using mirrord and get a sense of how the feature will behave, which gives us a lot of confidence."
β Reut Levy, Software Developer at @mondaydotcom
These are the results monday.com saw after adopting mirrord:
β±οΈ Environment setup dropped from 30 minutes to 10 seconds
π° Recurring cloud costs for dev environments eliminated entirely
π§βπ» A single shared cluster now serves 350+ engineers
π
New developer onboarding went from 3β5 days to about an hour
Read the full case study: metalbear.com/mirrord/case-sβ¦
Jake Page retweeted
Me and @jakepage91 Page ran into a problem related to DBs on our shared staging @kubernetesio cluster (it was all Jake's fault)
Learn what happened nextΒ π
Jake Page retweeted
Launch Week Drop 2/5: mirrord for Windows πͺ
mirrord now runs natively on #Windows. No WSL, no workarounds.
If you develop for #Kubernetes on a Windows machine, you can now run your code locally while it's connected to your cluster, same as your teammates on macOS and Linux have been doing all along with mirrord!
Get started here π
metalbear.com/launch-week/
Jake Page retweeted
This is what Itay Hilel from @daylight_ai had to say on how their team uses mirrord Preview Environments with AI agents to validate code and ship faster.
Read the full story here: metalbear.com/mirrord/case-sβ¦
Jake Page retweeted
Launch Week Drop 1/5: mirrord Preview Environments
With mirrord Preview Environments, you add a GitHub label to your pull request and get a shareable preview environment running in your existing staging cluster. No spinning up the entire application stack per branch.
This is a great way to collaborate with stakeholders like product managers, designers, and QA testers, and get feedback before anything is even merged.
Learn more about mirrord Preview Environments here: metalbear.com/launch-week/
Jake Page retweeted
AI coding agents are migrating from IDE assistants to unattended cloud workers (@cursor_ai's self-hosted agents, @cognition's Devin, @claudeai's managed agents).
Once they leave the developer's laptop, they need access to your real cluster's services, data, and queues to write and test code better. And they need this access without trampling each other or blocking on deploy cycles.
Our latest guide walks through the "agent pool with mirrord" pattern that companies like @mondaydotcom and @PodiumHQ are already running in production, allowing them to fully utilize the benefits of AI coding agents.
Learn more about it here: metalbear.com/blog/ai-agent-β¦
Jake Page retweeted
AI agents write code that looks correct. It follows the patterns, passes the linter, and ships.
Then it exposes your API keys. Or skips the auth middleware. Or calls a service that changed its response shape months ago.
Static analysis catches some of this. But it can't catch everything. We wrote up a four-layer approach that closes the gaps:
1. Skills: rules that shape generation before the code gets written
2. Audit commands: static checks that run regardless of how the code was written
3. Integration tests: run tests against real dependencies, not mocks to be sure things work
4. Preview environments: so reviewers see the feature running, not just the diff
Read how to wire them together: metalbear.com/blog/layers-ofβ¦
Jake Page retweeted
Replying to @threepointone
Thatβs what weβre building, minus the need to actually deploy - mirrord.dev
Jake Page retweeted
Weβll be at @CodemoMadrid tomorrow, 20th April! πͺπΈ
@jakepage91 from our team will be giving a talk where you, as the audience, interact with a real app in #Kubernetes until it breaks. The challenge is to reproduce the bug, debug it, and fix it without redeploying. Join to learn how to speed up your Kubernetes dev loop! β‘
If you're attending the conference, add the talk to your agenda here: conferences.codemotion.com/mβ¦
Is anybody using AI agents to write code that eventually lands in a Kubernetes cluster? Are you just doing regular post-CICD integration testing? Does the agent use pre-existing e2e tests?
Curious how you are effectively using AI code gen without breaking happy paths in k8s.
2
2
987
Jake Page retweeted
AI coding agents have joined your team, and they're not going anywhere. But without a realistic feedback loop, they risk quietly building on top of broken changes and burning through your token budget along the way.
Getting it right means giving the agent a way to know whether its code actually works. By running your agents with mirrord, they can execute pre-existing E2E tests against real databases, real queues, and real downstream services, all without deploying. The agent changes code, runs tests, reads the failure, fixes it, and moves on. One or two iterations instead of twenty.
Our new blog post walks through why this matters more than prompt optimization and how to set it up:
metalbear.com/blog/how-to-prβ¦
Jake Page retweeted
What better occasion than Easter for this very special and impeccably timed announcement: mirrord now supports @RabbitMQ queue splitting π°Β πΒ π°
Queue splitting lets you filter messages from shared queues so your local app only receives the ones meant for you, while the rest of your team keeps working undisrupted. We already supported #SQS and #Kafka, and now RabbitMQ joins the family.
If this made you jump for joy, hop into our docs and give your dev loop a giant leap forward: metalbear.com/mirrord/docs/sβ¦






















