Eric Minick

Leading horses to #DevOps.
Product and marketing work at me how I named Terraform before Hashi.

My posts are my own, not my employer's.

[Eric Minick @EricMinick]

Agents need to get sharper at using production telemetry to update code accordingly.

Excited for my @harnessio team as we're taking Cloud Cost data and updating IaC code (still w/ human review)

https://t.co/WpC9Wj0cuX

[Eric Minick @EricMinick]

Great session from Pritesh Kiri on LitmusChaos. Cool to see the momentum on the project, MCP server, and litmus-lite for local testing.

#kubecon

[Eric Minick @EricMinick]

We implemented some automatic budget management for our token'd AI users (everyone from developers to legal). I expected misery, but it's been pretty easy to work with and it looks like it saved a ton of money.

https://t.co/z7Kc48leS7

[Eric Minick @EricMinick]

Automated code review is important, but the downstream testing is critical.

Coding and review agents can’t ship safely on their own. The layers of defense we’ve built (or should have built) are:
A) still needed
B) must accelerate to keep up

[Ethan Lin @EthanLinAI]

AI code review bots catch fewer real bugs than you'd think.

CR-Bench (arXiv, Nov 2): 584 PRs, each with a known defect. 8 review agents on 4 backbones, plus Gemini Code Assist.

- Best systems: ~33% of known defects found
- Gemini Code Assist: 8.1%

Keep tests on the merge path.

[Eric Minick @EricMinick]

The Harness plugin has officially landed in the Google Antigravity marketplace. 🎉

There are 71 workflow skills, using Harness MCP. See into pipeline failures, deployments, and make pipeline YAML straight from chat, all without leaving Antigravity.

More: https://t.co/iJf1FEtpxi

[Eric Minick @EricMinick]

@PovilasKorop Deterministic for simple important things

Agentic can help when interpreting the results. A hard gate on severe security risks makes sense but when four tests are fixed and two new ones break, you’re in a world of judgement.

[Eric Minick @EricMinick]

@GergelyOrosz @Pragmatic_Eng If we ship 10x the content, the risk of change needs to drop 10x too or outage grows. Meanwhile SRE/DevOps teams are being pilfered to support AI rollouts. Interesting times.

[Eric Minick @EricMinick]

@GergelyOrosz @Pragmatic_Eng Very clear this is happening. Anyone who saw agile break everything downstream isn't shocked.

We'll work on review, then build, then deployments.. realize tests still stink and improve those. Then the tests are too long so we'll work again on selectivity. Repeat.

[Eric Minick @EricMinick]

52% of engineers report frequent or constant burnout.

71% are pulled into evenings or weekends by releases or production issues at least weekly.

Burnout is not just a people problem.

The cure seems to be fixing downstream automation...

New research: https://t.co/iYA62m3ZUW

[Eric Minick @EricMinick]

@linear @moofeez Very similar work to what Anthropic talked about a month ago. Agentic coding is breaking approximately everything downstream, in order. Reviews, then CI. CD is next.

Optimizing tests with a "test intelligence" is key for feedback and cost savings these days.

[Eric Minick @EricMinick]

With Claude and Cursor, we're seeing that shipping faster is easy.

Harder: Not lighting everything on fire when you go fast.

[Eric Minick @EricMinick]

Powerpoint gaining an edge on Google Slides because the underlying format is just XML and therefore easy for AI tools to work with is a fascinating case study.

Essentially everything benefits from an "as code" interface, and APIs are a pain in the ass / don't count.

[Eric Minick @EricMinick]

Top minds from TD, athenahealth, Amazon, Cursor, and State Street are digging into AI-driven software delivery at {unscripted} by Harness.

Explore the full lineup and secure your place: https://t.co/YpeyslQkHP https://t.co/FWmUmzMWEv

[Eric Minick @EricMinick]

It also points out that while GitOps will deploy your thing, it doesn't help you decide if you should. Yep... running tests and scans and making decisions still needs to happen. And it's probably happening in a pipeline. For a model, those tests are a bit different.

[Eric Minick @EricMinick]

GitOps wasn't built for Models? Harsh take.
https://t.co/RoSjPeF6Bz

The bits that are containerized are solid for GitOps, but LIKE EVERYTHING ELSE that's not everything which makes things messy for GitOps purists.

[Eric Minick @EricMinick]

Is there a database schema? Those get incrementally updated and rolled back (although we have good approximations for better).

The article points out that there are lots of things to version. That's correct. It's correct for lots of things.

[Eric Minick @EricMinick]

I like that @businessbarista emphasizes that verification is the hard part. He emphasizes testing, which is part of it, but for serious systems we need to validate change with performance checks, and business metrics. Those biz metrics mean automated feature management.

[Alex Lieberman @businessbarista]

https://t.co/v7Uphcl9Nw

[The New Stack @thenewstack]

DevOps scaling isn't just about automation — it's about architecture. Join @EricMinick & Stephen Cihak to learn how to build for the long haul on Feb 3, thanks to @harnessio.

➡️ https://t.co/mcJk5h07x9

[Eric Minick @EricMinick]

Recently talked to @thenewstack about reuse in #DevOps pipelines. Love the article!

https://t.co/OGnBqnRoJY

[The New Stack @thenewstack]

How to survive the #GitOps sprawl. 🚀 We’re diving into the data of enterprise teams who hit the "Argo Ceiling" and how they broke through. Join @harnessio's @EricMinick & @dewanahmed on Jan 20 for this free webinar.

https://t.co/Jic6tKaFEo