Vincent Hus

Following the great first OpenSRE community call on Discord, we’re hosting the next one tomorrow, Thursday April 16 at 17:00 UK time. discord.gg/A9J2dHdH?event… Come join if you want to jam on ideas, debugging, open-source, and building. See you there!

LLMs write “perfect” code in seconds. Engineers still have to debug the 3AM production outages. AI automated the easy part. We still own the fires 🔥. We’re building AI agents that actually matter: autonomous incident response. OpenSRE → github.com/Tracer-Cloud/o…

OpenSRE The open-source framework for AI SRE agents, and the training and evaluation environment they need to improve. Connect the 40+ tools you already run, define your own workflows, and investigate incidents on your own infrastructure. github.com/Tracer-Cloud/o…

@vaibhavupreti and I were discussing how to stay at the top of our game as engineers if the LLM models will improve by another 50% in the coming 6 months. How do we prepare ourselves? Taste, judgment, and decision-making are becoming more and more important. So the main thing to improve is decision quality: studying strong UX, proven system design patterns, and interaction models that work well in practice. From those studies and observations we've also decided to move from a static CLI to a more agentic one How are you adapting as an engineer?

Opus 4.7 is real Opus.4.6 and Opus 4.6 is 4.5 with the nerf

@RocketMan_mus Hey Zoe, good to see you again, how is your python programming going? You should also become an OpenSRE contributor!

Following the great first OpenSRE community call on Discord, we’re hosting the next one tomorrow, Thursday April 16 at 17:00 UK time. discord.gg/A9J2dHdH?event… Come join if you want to jam on ideas, debugging, open-source, and building. See you there!

What if one prompt turned your infra chaos into auto-resolved incidents? We built OpenSRE so you can build your own AI SRE agents - with 40+ integrations. Stop reacting. Start auto-resolving.

@open_sre @vaibhav__upreti @devesh__31 @0xcrackedDev @Ebrahim_Sameh_ @Abhishekcur @ra0van @skjtwts

@open_sre @vaibhav__upreti @devesh__31 @0xcrackedDev @Ebrahim_Sameh_ @Abhishekcur and everyone else who is currently in the chat (sorry if I don't have your X tag)

Updated discord link: discord.gg/zPRPtwx5gM

Hey all, we're kicking off our first OpenSRE jam session this week feel free to join on Discord discord.gg/9WHa6VQm?event… We’ll be building live, debugging real incidents, reviewing PRs, and shipping together hands-on. Bring: - something you’re working on or an outstanding PR - a bug / incident - or just join to watch and learn Would be great to have as many of you there as possible

Looking for open-source contributors Apache 2.0 Coding agents are everywhere. Production debugging still sucks. We're building an open-source agent that investigates productions incidents

Engineers in 2026 still getting wrecked at midnight manually grepping logs like it’s 2015. Wake the hell up! We just open-sourced your customizable AI SRE engineer - reads your logs, traces, metrics & runbooks, and finds root cause in minutes while you sleep. No more all-nighters. No more noisy alerts destroying your life. @open_sre is live. Try it today → opensre.com GitHub: github.com/Tracer-Cloud/o… What’s the worst alert that’s ever ruined your night/weekend? Drop it below 👇

@sarafoleanu 100%

@vincent_hus Agree SWE-bench doesn’t tell the real story. Production complexity is where this gets scary. Models chaining real IAM misconfigs to security group gaps to flat subnets. That’s the benchmark that matters for defenders

They're benchmarking the new claude mythos on SWE-bench Meanwhile OpenSRE is building a significantly harder distributed systems version of it In a while, the real test for these frontier models might actually shift from SWE-bench to something closer to production-grade complexity The benchmark game is about to level up for everyone

Great write up about OpenSRE's synthetic test cases The implementation of reinforcement learning cases critical to get agent reliability right dev.to/kalio/how-ai-e…

The autonomous agentic future feels like the mobile data of the 2010s OpenClaw promised 30-second ECR redeploy Reality: 10+ minutes of "And...", "Yo", "Hey Vincent!" just to get another cryptic exit code Anyone else fighting these OpenClaw is not responsive issues?

@LeinRune @LeinRune thanks for pointing it out, should be fixed now!