DevOps Social Media and Updates
DevOps Social Media and Updates
DevOps Pulse
EMAIL DIGESTS
Daily
Every morning
Weekly
Tuesday recap
EMAIL DIGESTS
Daily
Every morning
Weekly
Tuesday recap
I’ve started muting all the accounts pushing hot takes on ai 🤖 so I can focus on getting things done. If it’s not contributing towards accomplish something not really helping. Here’s what I’ve gotten done or trying to get done with vibe coding with a human in the loop: Networking is still the most complicated thing on AWS - if you bother to set it up correctly. I’ve spent a lot of time creating a script that recommends a default configuration for a new organization in an AWS account. Then it allows the user to change it. I’m at the point of using that configuration to deploy the VPC and related resources. Working through issues. Also write a post on DNS AI leaks since it was on my mind and a couple of scripts to mintier did that - one for my Mac and one for an EC2 instance. Which brings me to my prior post again - why is an MDNS service running on my EC2 instance? I wrote a script to kill that and all the other leftover SSH sessions hanging around when I get disconnected from SSH - which is happening a lot lately. Now when I reconnect I can run the script to stop all that and either it any rogue leftover agents. I also am working towards the goal of the Yubikey to kick off a job solution. While setting up accounts the “save another way” option for MFA was not working temporarily. I set that aside and when I came back again it worked. I ended up writhing a blog post on where credentials are stored when you configure a passkey and how they can be transferred to the cloud depending on what option you choose. After sufficiently satisfied with my target Yubikey registration process, back to the environment implementation. I added SCPs and realized I needed to add role permissions. At this point I also realized you cannot put conditions on all actions that can affect an SCP attached to the root OU. Did the best I could and added an item to my #awswishlist at https://t.co/rvPGzaBSK3. Happy to see more people are using that. The SCPs limit access for roles in an environment so they cannot do things outside their own environment amount other things. Ok back to the VPC hold up. Last big thing before I can really start testing auth and job invocations.
Okay this might be totally obvious but you can just tell Claude Code to double your test coverage and it will actually do it. Will need a lot of review, but if you’re just getting started with adding tests to your...
I am starting to believe that the most interesting part of AI development is not the models. It is the workflows around them. Agents coordinating work, orchestrating tasks, managing context. Feels a lot like the early days of DevOps when we started wiring...
Spent two weeks benchmarking Pilot on Terminal Bench 2.0. Ran 500+ tasks across 15 experiments. Built analysis pipelines. Measured variance. Compared agent behavior across pass vs fail runs. The fix that moved the needle? Removing one env var that forced maximum thinking...
In 48 hours our AI agent simulated 6 months of activity in the app. Anthropic dropped a 1 million context window last weekend and we put it to work. No human intervention. The results: • 30% faster build times • 50% reduction in...
GStack just shipped Windows support. Thanks to all my Windows users for bearing with us. There was a Bun runtime error on Windows and now it falls back to Node.js. Big thanks to the GStack contributor community - sorry it...
RT Your new backlog shouldn't "add AI." It's redesign feedback loops so models, pipelines, and platforms learn together. Think: model drift to incident response, feature flags to guardrails, SLOs to AI behavior. #AI #DevOps #Agile @Star_CIO https://t.co/7dcoLIKa0K
20 steps to improve API performance ⚡️ DB + code: indexing, query caching, pooling, efficient algorithms Caching: Redis/Memcached, HTTP cache headers, CDNs Transport: GZIP/Brotli, HTTP/2–3, keep-alive, TCP tuning Scale: pagination, async processing, load balancing Reliability: rate limiting, timeouts, proper errors Ops: monitoring/profiling, versioning UX: smaller payloads, better...
Agentic AI can turn a single architect into a "full-stack squad" in minutes, but it can also create black-box apps that no one can safely extend. #AI #DevOps https://t.co/p18hdtdbZn
If only my Kubernetes pod could scale to zero. That'd be great for staging/test environments or irregular production workloads. Oh, that's coming in Kubernetes v1.36 after sitting in alpha for years? Sweet. https://t.co/klzNA6Hs0X https://t.co/Yee1mpTLEh
This direction is very meaningful, since translating intent into executable tests can reduce friction in QA workflows and let teams focus more on quality thinking rather than repetitive implementation.
Node 18 + ClaudeCode 2.1.72 is a cheat code 😉 We benchmark Pilot on Terminal Bench 2.0. 89 real coding tasks, Opus 4.6, Modal containers. Ran 10+ full experiments over two days. The CLI tool version matters more than prompt engineering, effort...
Fragments: code review isn't just catching bugs, what role for observability in agentic programming, what we lose with GPS over maps https://t.co/f2tfkLanN4
"Formal methods and test suites are flight simulators. Production is flying the actual plane. Observability is how you fly it." -@mipsytipsy Important post from Charity on why you must treat production as more than a place you go to fix...
Most people say they “know CI/CD”… until they get asked to explain it in an interview. Can you clearly break down: CI vs CD Pipelines end-to-end Artifacts + versioning Rollback & deployment strategies Secrets + security That’s where most candidates fold. DevOps isn’t about memorizing tools — it’s...
One of my most painful realizations using agentic systems is that they treat all work-in-progress as ephemeral. They’ll gladly reset your branch or dev database to “try something”, because they expect to write working code eventually. Never manually add code while...
Honestly, @Trae_ai on Linux is exactly what I needed, an intelligent assistant that understands my workflow and speeds up development in my preferred environment. With Minimax M2.7 now built in, I can take on even more complex tasks effortlessly. https://t.co/1kmZKxIeBt https://t.co/S5QDeHBm04
I added a /codex skill that does plan and code review in Claude Code and it has found so many more bugs and helped with far more elegant architectural designs And tomorrow I will launch native Codex support so you can...
Kiro + Opus 4.6 🤖 Struggling with - cross account lambda layer permissions and recognizing they are needed - CORS issues - maintaining correct variables and ARNs for multi-region deployments - Repeated unbound variables in complex bash scripts - making up variables that don’t exist -...
Reducing the QA phase to one day to make up for delays in the design phase is a massive trap.
A new release from @cursor_ai turns its internal security agents into reusable templates teams can deploy themselves. For developers, that could mean continuous security checks — but it raises questions about trusting AI to review code. 🔗Story here: https://t.co/yp2DZC9LrL https://t.co/ugKk1HaKki
We gave every person at Wynter 1 week to build their own AI agent. Budget of about 4 hours each. Today was demo day. Everyone presented what they built. Most had never built anything like this before. Some had zero technical background....
The most underrated AI role right now: DataOps Engineer. Not the ML engineer. Not the data scientist. The person who designs automation and testing infrastructure that makes everyone else dramatically more effective. Infrastructure that runs without you. That's the whole job. https://t.co/Cng5iC1BEB
For those running AI/ML models on Kubernetes, do you feel pinned to one region? Not great. We just previewed the @googlecloud multi-cluster GKE Inference Gateway. Scale inference workloads across clusters, even across regions. https://t.co/t6vL4a7ZEH https://t.co/MTpFrlroKp
Last 6mo of @vercel build & deploy perf ships: ① 𝙼𝚊𝚛 𝟼 - 𝟷.𝟸𝚜 - 𝟹.𝟽𝚜 𝚏𝚊𝚜𝚝𝚎𝚛 🆕 Deployment step now 15% faster ② 𝙹𝚊𝚗 𝟸𝟹 - 𝟼𝟶𝟶𝚖𝚜 - 𝟻𝚜 𝚏𝚊𝚜𝚝𝚎𝚛 Faster deploys with improved function caching ③ 𝙹𝚊𝚗 𝟷𝟺 - 𝟸.𝟾𝚜 - 𝟷𝟸𝚜 𝚏𝚊𝚜𝚝𝚎𝚛 Reduced...
So this turned into build an environment to run in. I wrote a script to deploy all the resources with guidance: Tested and deployed so far: Org root user and role Root policies Root permission boundary ou accounts - iam - KMS - dns - jobs - work - backup...
If you haven't already, tail your server logs and filter by requests that take longer than 500ms. Copy paste it all into Claude and say "pls fix". Instantly faster app.
Which tool is best for beginners starting DevOps? 👇 1️⃣ Docker 2️⃣ Linux 3️⃣ Git & GitHub 4️⃣ Kubernetes
Perk up when the DORA team does research into software delivery. The surveyed Google engineers to understand how AI tools impacted their work, and where they struggled. The advice is very actionable. https://t.co/PnaUMyiBLp https://t.co/mgU4WJb8SL
As someone who has burned more hours than he wants to think on server administration, a *lot* of the cruft of it gets transformatively easier with LLMs, and in lieu of an hour doing deferred maintenance you can spend an...
RT Multicloud != "just more clouds." It's divergent APIs, IAM models, pricing, and PaaS semantics across AWS, Azure, GCP, Oracle, and others. GenAI introduces a translation layer for configs, code, and policies. #MultiCloud @Star_CIO https://t.co/vBzM21vM14
If your MCP server doesn't enforce data scopes, PII controls, and environment isolation, you're not "experimenting with agents" - you're opening side doors into production. #AI #DevOps #MCP https://t.co/7dcoLIKa0K
A real AWS Data Science pipeline looks like this: Raw data → S3 ETL → AWS Glue Query → Athena Training → SageMaker Deployment → Endpoints Monitoring → CloudWatch Add streaming with Kinesis and orchestration with Step Functions, and you have a full production ML platform. This is...
If you followed my journey to try to build a batch job framework (below) for like three years well, here’s what I got done vibe coding 🤖 as a chaperone for naughty AI agents and chatbots in two weeks. To...
Agents fail at backends because they lack context. @insforge_dev V2 fixes this via a semantic layer & MCP 👀 → Connects Claude/Cursor directly to your backend → Auto-configs Postgres DBs, Auth, & S3 Storage → Deploys edge functions seamlessly Free and open-source 🧵↓ https://t.co/BZmifQ1pqL
I don't want to read a giant troubleshooting guide, even a great one like this for GKE clusters. https://t.co/13E0PvzrMI Feed this as context into your agentic CLI (or use our @googlecloud Docs MCP) and send your agent down the right path faster.
Have you ever worked in a company where developers get stucked in “ticket ops” this is meant for platform engineers, where developers need the manual approvals as a platform engineer you will be tasked to develop a system or product...
Deploying a model is harder than training it. 🚀 Here’s a simple ML → Production pipeline on AWS: Train model → Build API → Dockerize → Push to ECR → Deploy with Lambda → Serve predictions. Notebook models don’t create impact. Production models do.
Wow. @GarryTan (@ycombinator's CEO) just dropped the ultimate cheat code for software engineers. 🔥 He just open-sourced gstack, his personal toolkit that transforms Claude Code from a basic chatbot into an entire virtual engineering department. Instead of asking Claude to "build a feature"...
Your new backlog shouldn't "add AI." It's redesign feedback loops so models, pipelines, and platforms learn together. Think: model drift to incident response, feature flags to guardrails, SLOs to AI behavior. #AI #DevOps #Agile https://t.co/7dcoLIKa0K
OK, well. I ran /autoresearch on the the liquid codebase. 53% faster combined parse+render time, 61% fewer object allocations. This is probably somewhat overfit, but there are absolutely amazing ideas in this. https://t.co/dpEJw7NpL4
Impressive work from @QodoAI Their extended mode does a great job finding more real code issues without adding lots of false alarms. That means code reviews stay accurate and less noisy. If your dev team deals with messy PRs, this could help you...
Created my own custom AI chat bot trained on my site data using Cloudflare AI Search/Workers and Cloudflare AI Gateway which has firewall, Guardrails option feature. Seems could be quite expensive to enable 🤔
Writing crystal-clear instructions for machines is the new 10x dev skill. The most important file in your repo is no longer your code. It’s your `CLAUDE.md` 🔥 Top devs are using this file as an onboarding doc for AI. Instead of typing lines,...
AI won't fix broken DevOps. It amplifies whatever culture, architecture, and practices you already have. If you're shipping chaos today, AI will just ship it faster. Time for leaders to refactor ways of working, not just toolchains. #AI #DevOps https://t.co/7dcoLIKa0K
System-first o11y has its function, but more and more devs are realizing that effective debugging requires a session-first approach with a single timeline that correlates: • User interactions • Console errors • Network req/res • Backend traces and spans
This AI chatbots 🤖 do while vibe coding: I have a complex bootstrap script to set up an environment for my batch job framework. It needs to create accounts with a role in the root account. And that is all it...
In June 2025 @darraghcurran announced our 2X initiative (link in thread). Below @brian_scanlan shares where we're at today (~90% of PRs created by Claude Code). It's exciting to see all this work come to fruition.
We're racing to build AI agents, but many are still stuck in the "app testing" era. Gen AI QA must validate non-deterministic outputs, data readiness, and real-world risk - not just pixel-perfect screens. #AI #DevOps #QA https://t.co/1JJW0fGZaG
If I had to scale an app from 0 to 1 million users, here's what I'd do (step by step): There are different ways to scale & this is just one of them. The best solution & numbers will depend on your needs. Keep...
