The speed headline is real. The inspection story is better.
Homebrew 7.0.0 overlaps more download, preparation, and installation work across install, reinstall, upgrade, and Brewfile batches. It also reuses parsed API data on warm runs while continuing to verify signatures on every load, and trims startup subprocesses.[1] Faster package plumbing is welcome. Faster uninspected package plumbing is merely a quicker surprise.
The release pairs speed with new inspection surfaces: brew install --dry-run previews formulae and casks; brew deps --brewfile exposes what a Brewfile pulls in; brew vulns --brewfile checks known vulnerabilities against Homebrew’s new advisory database; and brew doctor --json turns workstation diagnostics into structured output.[1]
The grown-up package-manager feature is not “install faster.” It is “show me the blast radius before the fans spin.”
Turn onboarding into a reviewable patch.
A new developer often receives a README line that mutates the machine and a shrug about what it installs. Homebrew’s own installation documentation says the default prefix matters for bottle use, explains the initial privileged setup, and warns that configured Git mirrors receive the same trust as Homebrew itself.[2] Those are policy facts, not footnotes.
The better onboarding pull request includes the Brewfile, its expanded dependency map, the target architecture and support tier, the vulnerability report, and a diagnostic receipt from the actual host. Reviewers can then ask useful questions: why is this cask present, which tap owns it, does it need a service, does the Intel machine have a viable binary route, and what is the removal plan?
Sandboxing shrinks damage. It does not mint trust.
On macOS, Homebrew 7 tightens cask sandbox restrictions. On Linux it moves from Bubblewrap to Landlock where the kernel supports it. The release also says package builds are migrating toward a networked fetch phase followed by an offline install phase with read-only caches.[1] That separation is exactly the kind of boring boundary developer tooling needs.
Homebrew is equally explicit about the limit: tap trust remains primary; sandboxing cannot make untrusted software safe; applications eventually run with the user’s privileges; vendor .pkg installers may run outside the sandbox and ask for sudo.[1] A green vulnerability scan is not a provenance verdict, and a sandbox is not a moral purification chamber.
Machine support belongs in the receipt.
Homebrew 7 drops macOS 10.15 and moves Intel Macs to Tier 3, with no routine new bottles and an announced September 2027 end date. The official installation page now lists Apple Silicon with macOS 15 or later as the supported macOS path while describing Intel as Tier 3.[1][2]
The Hacker News discussion quickly became a referendum on old Macs, Linux package boundaries, MacPorts, mise, and Homebrew’s supply chain.[4] That is useful discovery context, not release authority. The practical rule is simple: write the architecture, OS, prefix, support tier, and fallback package manager into the workstation contract before the upgrade, not after bottles disappear.
Four commands before the pour.
- Preview: run
brew install --dry-runfor the proposed direct package set. - Map: run
brew deps --brewfile; review taps, casks, services, language tools, and transitive weight. - Scan: run
brew vulns --brewfile; record skipped coverage and distinguish “no finding” from “no risk.” - Diagnose: run
brew doctor --json; attach machine, prefix, architecture, OS, support tier, and warnings. - Then pour: install from a reviewed revision and preserve the before/after receipt.