Mike McQuaid

4.6.0 Today, I’d like to announce Homebrew 4.6.0. The most significant changes since 4.5.0 are opt-in concurrent downloads with HOMEBREW_DOWNLOAD_CONCURRENCY, preliminary macOS 26 (Tahoe) support and a buil...

🍺 Homebrew 4.6.0 is out!

Opt-in concurrent downloads (`HOMEBREW_DOWNLOAD_CONCURRENCY`), early macOS 26 (Tahoe) support, a built-in `brew mcp-server` and lots of other fixes.

Run `brew update` to get it.

Full details at: brew.sh/2025/08/05/h...

Joel Spolsky's "Smart and Get Things Done" remains best summary for engineering talent.

No-one gives a shit how you did in school or what your IQ is if you can't actually ship.

No-one cares how fast you ship if you are stupid about how you do it.

Move faster, intelligently.

"Well, will you decompose your monolith into microservices just because all your friends are doing it?"

Where these tools really shine is for cases where the realistic alternative is not "an engineer will do it" but "no-one will bother doing it at all".

There's many scripts I wrote this year that I just wouldn't have bothered with before. Time spent to saved wouldn't make sense.

"ChatGPT cannot write better code than a mid-tier engineer"

Ok cool I guess I'll ask the mid-tier engineer who is willing to work for £20/month and be on call 24/7/365 to answer my queries in under a minute to write that script for me instead.

For example, if you do `git commit` and then `git push`: commit locally and skip only the push to build confidence everything else is working.

Another example, for release workflows: tag in the workflow and only `git push` the tag after you've confirmed everything pasts tests, builds, etc.

GitHub Actions conditional logic best practice: if you have code that runs only on a push to `main`/`master`, on a release, schedule, etc. run as much of it as possible every time.

Some of the best advice I ever got from seasoned OSS leaders was to stop doing work that others on the project could do instead. Step back, let others step up and focus my energy on higher impact problems.

This advice proved even more applicable as CTPO of @workbrew.bsky.social, too.

Why Open Source Maintainers Thrive in the LLM Era At the time of writing (June 2025), the prevailing view in the software industry is that LLM-powered AI is either completely useless or will imminently destroy all software engineering jobs. As you mi...

I discuss building this PR in particular with LLMs and my approach a bit in my recent post about why open source maintainers thrive in the LLM era: mikemcquaid.com/why-open-sou...

Add `brew mcp-server`: a MCP server for Homebrew. by MikeMcQuaid · Pull Request #20041 · Homebrew/brew Add a new brew mcp-server command for a Model Context Protocol (MCP) server for Homebrew. This integrates with AI/LLM tools like Claude, Claude Code and Cursor. It currently supports the calls need...

I just released @brew.sh 4.5.5. Normally patch releases aren't particularly notable but this one includes an MCP server for Homebrew (`brew mcp-server`): github.com/Homebrew/bre...

Let me know any thoughts or feedback!

And I'm seeing similar extremes in the Research Community when it comes to the usage of AI. Folks wholesale discounting it because of a hallucination, and other folks being like "we use synthetic users and don't talk to humans anymore and the results look credible!" (gell-mann amnesia effect).

Mike! I thought this whole post was SO SMART and really liked your points about the gell-mann amnesia effect (which I had never heard of!) and your points about asymptotic progress. I had to ask ChatGPT to explain your big words to me, but I really liked your thoughtful take 💜

Aw, thanks Grace! Hope you're well!

Why Open Source Maintainers Thrive in the LLM Era At the time of writing (June 2025), the prevailing view in the software industry is that LLM-powered AI is either completely useless or will imminently destroy all software engineering jobs. As you mi...

Read full thoughts in a post on my site: mikemcquaid.com/why-open-sou...

Why will open source maintainers thrive in the LLM era?

> They have honed the skill of rapidly reviewing large volumes of unfamiliar or newly contributed code. This skill is useful, perhaps even essential, for effectively leveraging LLM output.

Interested? Check out the post:

Read the full piece here: opensource.org/maintainers/...

On challenges:

> Sometimes, things can get darker still with things bleeding into personal abuse and harassment. I once had someone say they were going to turn up and harass me at a conference talk I was giving. I’m lucky enough to not be threatened by this but: it’s completely unacceptable.

(I have a WIP post for going in more detail on what I think about LLM AI and OSS, stay tuned!)

On AI:

> I think LLM tooling has the potential to be positive for open source but the jury is still out. To be really effective: you need to ensure you do extensive review of any AI generated output. Who are some of the best people in the world at extensive code review? Open source maintainers.

Wrote a piece for @opensource.org based on my experiences in open source over the last 20 years (!).

Co-founders John Britton & @mikemcquaid.com joined @johncodes.com and @bizza.pizza on @heavybit.com Open Source Ready podcast to share the story of Workbrew. They unpack their journey in dev tools, the evolution of Homebrew, and what’s next for developer environments.

🎧 buff.ly/hqUKLeT

4.5.0 Today, I’d like to announce Homebrew 4.5.0. The most significant changes since 4.4.0 are major improvements to brew bundle/services, preliminary Linux support for casks, official Support Tiers, Tier 2...

Read the release notes at brew.sh/2025/04/29/h... and discuss on Hacker News at news.ycombinator.com/item?id=4383....

Proud to release Homebrew 4.5.0 today. The most significant changes since 4.4.0 are major improvements to brew bundle/services, preliminary Linux support for casks, official Support Tiers, Tier 2 ARM64 Linux support, Ruby 3.4 and several deprecations.

Proud of this one. Following a Shape Up process to get new features to customers every ~8 weeks. The velocity of shipping is incredible over here.

How and why I interview engineers for Workbrew In the last two years building Workbrew (a remote-first, enterprise Homebrew startup) I’ve hired 5 engineers (and a hybrid PM/EM). This has been my first time being a “hiring manager”. This post expla...

Read more in mikemcquaid.com/how-and-why-...

I designed the initial engineering hiring process at Mendeley and interviewed the first ~5 engineers there. At @github.com, I interviewed ~50 engineers. Through these, and being interviewed ~15 times myself, I had a good idea of what I loved/hated to design things accordingly.

In the last two years building @workbrew.bsky.social (a remote-first, enterprise Homebrew startup) I’ve hired 5 engineers (and a hybrid PM/EM). This has been my first time being a “hiring manager”.

It's really underrated how much, when building a CLI, it's worth leaning on integration/regression tests that actually run the tool rather than "test" the code in some unrepresentative test environment.

These tests have saved me on Homebrew and other projects more times than I can count.

and discuss on Hacker News if that's your thing in news.ycombinator.com/item?id=4360...

Two Years Building Workbrew, a Remote-First Enterprise Homebrew Startup TL;DR: I left GitHub after 10 years to cofound Raise.dev. After an early failed experiment building IoT developer tools, we pivoted and built Workbrew: an enterprise-friendly layer over Homebrew, a pr...

“Minimum viable management” is helpful. Zero management is not.

Read the rest in: mikemcquaid.com/two-years-bu...