As a stealth ceo of a profitable SaaS. This is a nice reminder for my company to wind down its relationship with OpenAI. I have no doubt Anthropic will eventually become evil but at least they have a backbone today.

Goodbye Sam.

Edit: Also, referring to the DOD as the Department of War is cringe.

In September 2025, Executive Order 14347 authorized the usage of "Department of War" as a secondary name, which is now preferred by the department.

Cool

Local inference might a better bet for you.

Killing engineer teams? Hyperbole thread titles need to be killed. I find github actions to be just fine. I prefer it to bitbucket and gitlab.

Yeah I was wondering how Microsoft is okay with Github murdering people but then was let down by the article.

I clicked the article thinking it was about GitLab. Much of the criticism held true for GitLab anyway, particularly the insanely slow feedback loops these CI/CD systems create.

Can't blame gitlab for team not having a local dev setup.

You can though. GHA and Gitlab CI and all the others have a large feature set for orchestration (build matrices, triggers,etc.) that are hard to test on a local setup. Sometimes they interfere with the build because of flags, or the build fails because it got orchestrated on a different machine, or a package is missing, or the cache key was misconfigured, etc.

There are a bunch of failures of a build that have nothing to do with how your build itself works. Asking teams to rebuild all that orchestration logic into their builds is madness. We shouldn’t ask teams to have to replicate tests for features that are in the CI they use.

Indeed there are. But you iterate on local and care about CI once everything is working in local. It's not every tuesday I get CI errors because a package was missing. It's rare unless you're in those 1000-little-microservice shops.

It is rare for our run of the mill Java apps to however, we notice it with:

Integration of code quality gates, documentation checks, linting, cross architecture builds, etc.

Most of this can be solved by doing the builds in a docker image that we also maintain ourselves. Then what remains is the interaction between the ci config for matrices, the tasks/actions to report back quality metrics, the integration with keyvaults to obtain deploy time secrets, etc.

Then there are the soft failures, missing a cache key causing many packages to be downloaded over and over again, or the same for the docker base images, etc.

We fix this for our 1000+ microservices, across hundreds of teams by maintaining a template that all services are mandated to use. It removes whole classes of errors and introduces whatever shenanigans we introduce. But it works for us.

If GHA, Azure Pipelines, etc., would provide a way of running builds locally that would speed up our development greatly.

Until then we have created linting based on CUE to parse the various yamls, resolving references to keystores, key ids, templates, etc., and making sure they exist. I think this is generic enough to open source even.

I send an email to myself. Monday todo, Tuesday todo etc..

I pair crons with healthchecks.io

This is cool and after learning of this I'm tempted to convert my crons. I currently use systemd for running a redis queue and it's worked great for years.

How does this compare with railway?

We are trying to be more focused on just Docker and our pricing model is different. At sliplane you pay a flat fee for your resources and never more, at railway you pay exactly for the resources you use. If you have high utilization sliplane should be cheaper, but that depends on your usecase!

And we don't get to wine why?

The solution is to have a cheap phone that you sync with your authy. I had the trifecta with my desktop, now I just have one backup device because of this change.

Yeah its not sexy, but it does work and on Linux too!

Yeah its super convenient. I do keep an old backup phone around just in case and occasionally turn it on to sync, but also have the desktop app.