Joe Terlecki (@cloud_custodian) profile picture

Staff SRE, Neovim BTW, Recovering Linux User.

YinzerBurgh
Joined July 2021
Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)Photo by Joe Terlecki (@cloud_custodian)
Uh oh @aarondfrancis vibes all over again. Soon you will beg Twitter for a job and then release a database course course!
I'm so excited to announce that I've joined @PlanetScale as a content creator / educator! This is going to be huge. Lots of great stuff coming up. Stay tuned...
2
112
As much as I hate vscode. It's remote development is the goat
Now I'm running VSCode because both Zed and JetBrains are broken on remote.
1
147
Bitbucket has been really trying to modernize and I'm here for it. @Bitbucket @Atlassian Please add template repositories to your platform and I promise you will start stealing more market share. And it will make me happy
1
65
This is why I think ruby devs live the most chill lives. 1. Love the magic 2. Ignore the types 3. Ship features 4. Go home
Replying to @thekitze
Who gives a fuck about the quality of the code? Does it provide value? “Omg, I’m so glad I’m paying a subscription to this tool, the code base is soooooo clean and perfect” Said no consumer ever.
1
72
And the final icing on a cake. We migrated away from kubernetes 2 years ago. Best decision we ever did and now we focus on what truly matters delivering value. So you don't need kubernetes to make 200k?
Not because he knows the best way to configure a deployment yaml or a node constraint. He knows how to work with the teams with a shared goal. B be a problem solver. Be nice and show up when it counts then you'll get paid
61
Not because he knows the best way to configure a deployment yaml or a node constraint. He knows how to work with the teams with a shared goal. B be a problem solver. Be nice and show up when it counts then you'll get paid
I'm going to top this all off. Depending on a seniority, I won't expect all of this. I look for aptitude and the ability to learn and be curious. My right hand man came in not knowing exactly what a pod was and 2 years later he's an invaluable resource to the dev teams
1
97
I'm going to top this all off. Depending on a seniority, I won't expect all of this. I look for aptitude and the ability to learn and be curious. My right hand man came in not knowing exactly what a pod was and 2 years later he's an invaluable resource to the dev teams
I want you to be able to understand application code and architecture at dec level get in a code and debug with an engineer on why an AWS API is behaving a certain way. None of this stuff has to do with the Manifests and scheduling. Those are the easy problems in my opinion
1
73
I want you to be able to understand application code and architecture at dec level get in a code and debug with an engineer on why an AWS API is behaving a certain way. None of this stuff has to do with the Manifests and scheduling. Those are the easy problems in my opinion
I'm also tired of conflating the devops being a kubernetes engineer. The amount of times I had to worry about c groups for k8s clusters was less than 1. I want my engineers to be advocates and champions for good architecture, security and resources for our development
1
90
I'm also tired of conflating the devops being a kubernetes engineer. The amount of times I had to worry about c groups for k8s clusters was less than 1. I want my engineers to be advocates and champions for good architecture, security and resources for our development
Also, operations and devops engineers tend to try to make their tools and systems perfect out the gate which in turn slows down development and product. Good engineers know how to iterate and work your way to good enough to meet the needs and build from there as needed
1
1
108
Also, operations and devops engineers tend to try to make their tools and systems perfect out the gate which in turn slows down development and product. Good engineers know how to iterate and work your way to good enough to meet the needs and build from there as needed
You just need to know enough to find your own way But if you don't understand systems and how they can fail and things Cascade. I want someone who is cool during critical outage the day our customers run payroll. Someone with intuition to peel the layers while the fire is hot
1
84
I also look for strong signals of good software engineering practices You can be the best devops engineer CLI guru in the world But if you don't know how to manage code, contribute to internal tools, when to move from bash to python to go etc When you don't need perfect
You just need to know enough to find your own way But if you don't understand systems and how they can fail and things Cascade. I want someone who is cool during critical outage the day our customers run payroll. Someone with intuition to peel the layers while the fire is hot
1
50
You just need to know enough to find your own way But if you don't understand systems and how they can fail and things Cascade. I want someone who is cool during critical outage the day our customers run payroll. Someone with intuition to peel the layers while the fire is hot
This is why for my teams I don't hire "devops" engineers Give me a decent baseline of Linux, networking, docker, and python I'm concerned with systems engineering, architecture and trade offs I could care less how many obscure CLI args you know That's how you get 200k+
1
128
This is why for my teams I don't hire "devops" engineers Give me a decent baseline of Linux, networking, docker, and python I'm concerned with systems engineering, architecture and trade offs I could care less how many obscure CLI args you know That's how you get 200k+
The difference between $80K and $170K in DevOps: $80K — can build a CI/CD pipeline $170K — understands Linux, networking, containers, and Kubernetes end to end Most people stop at pipelines and wonder why they're underpaid.
1
1
116
I did not have getting Warhammer pulled on my bingo list for 2026...
30
Llms have been writing all my terraform. I'm ok with this
16
Static type advocates mass calling a dynamic language founder "stupid and dangerous" is wild considering: This dangerous dumbass bootstrapped a 100m business and continues to provide gifts to the world Y'all are ridiculous
What a perfect illustration of the type brain in its most rabid form. Catastrophizing the use of dynamic typing to the point of death. Fascinating state of mind.
44
my ADHD is all over the place. Being born into typed compiled languages and microservices and cloud native made me ignore the Elegance of Ruby and rails Been messing with both @bunjavascript and @rails over holiday break for a elegant full stack back end.. Ruby is such a fun language to write. Y'all been lied to. Rails 8 and modern Ruby is dope. Thanks @typecraft_dev and @dhh for helping me see the light
2
59
Find you someone who stares at you the way Davids agent does
One of the scariest and most fun part of coding with an agent is that it never tells me that I can’t do something…
1
30
A family member uses salon POS software valued at tens of millions. It's PHP from 2005. Still works. Still prints money. Meanwhile, we're told you need React, Redux, React Query, Tailwind, Vite, Docker, Kubernetes, and Terraform just to render text and save JSON. @dhh nailed it, we've lost sight of what made software fun to build. We've mass-convinced ourselves that shipping requires suffering. The complexity isn't the product. It's the cope. I avoided frontend for years because of this. Recently picked up @bunjavascript, HTML templates(pug, ejs) and HTMX keeping context in my head instead of spread across 47 config files. It's been a breath of fresh air. Bun made JavaScript fun on the backend again. Turns out you don't need Next.js and the hype machine to build something real. Build something that works. Ship it. Stop cosplaying as a platform team.
2
124
The amount of reach these engineering profiles/posts get that are obviously AI driven is crazy. Felt for it for a few days only to see the formula. Post with AI, Repost real industry through leaders with AI input. Reply with AI. Smdh
A durable execution isn’t just “code that survives crashes.” It’s an execution that is split across time, with gaps where the process may not exist at all — and then later resumes. That makes it fundamentally similar to a distributed system, just along the time axis instead of the network axis! When an execution is durable, it can be: - paused, - checkpointed, - crashed, - restarted on a different machine, - resumed hours or days later. Between those steps, memory is gone, stack frames are gone, threads are gone. The only thing that crosses the boundary is persisted state. That’s exactly the same constraint you face when sending a message to another service over the network. This is why durable systems force you to think in messages, state machines, and idempotency. Once an execution spans time, you can’t rely on in-process assumptions like “this line runs once” or “this function won’t be retried.” Retries, duplication, and partial progress become normal — just like in distributed systems. It also explains why durable workflows resemble event-driven architectures. Each step must be: - replayable, - restartable, - side-effect aware, - safe to run again. If a step isn’t idempotent, a crash turns into corruption. If state isn’t explicit, recovery is impossible. The execution must be decomposed into transitions that can survive being interrupted at any point. This is why tools like workflow engines, durable queues, and databases feel so similar to distributed coordination systems. They aren’t “just persistence layers” — they are time-distributed runtimes. Once you see this, a lot of design choices click: - why logs beat mutable state, - why replay is powerful, - why exactly-once is an illusion, - why “at-least-once + idempotency” keeps showing up. Durability isn’t about surviving crashes. It’s about accepting that your program no longer exists continuously, and designing it the same way you’d design a system that runs across unreliable machines — except the unreliability is time itself.
1
37