Roadmap

What exists, what is next

This page separates what is built, what still has to be checked on real Macs, and what is only planned. It says what is not done as plainly as what is.

Status as of 2026-10-15. Version: 1.0 (not yet released). Hedroom is not released. Code and tests exist for everything under "Built", and much of it has only been run against fake Macs so far. Timing depends on things outside the code (signing accounts, beta testers), so the windows below are the original plan and will move. Release notes will be on the changelog.

Where we are

Six phases, in order. A phase ends with a check that includes something observed on a real Mac, not only passing tests.

Make 1.0 safe and honest

Done

Fixes from an independent code and behaviour audit: Hedroom never signals Apple processes hiding inside a third-party app, paused apps cannot stay frozen after a crash, cleanup goes through the same safety checks as any other action, deletions need your permission each time, public copy matches the code.

Finish the 1.0 features

In progress

Every 1.0 screen and flow is built: error states, receipt detail and Put back, library updates, the weekly report, the Today screen, an accessibility pass with zero findings left in the audit rig, and all text in a string catalog (English only for 1.0).

Left: Performance with windows open is fixed on one Mac, but the popover, Settings, the other tabs over 10 minutes and a 30-minute leak run are not re-measured.

Check it on real Macs

Not started

Everything above is tested with fake Mac sources. This phase repeats the risky parts on real hardware and records the results.

Left: See "Checked on real Macs" below.

Release engineering

Waiting on accounts

Build, sign, notarize, staple, package the DMG, publish the update feed, and prove an update from one build to the next. The scripts are written and tested in dry-run mode.

Left: Needs a Developer ID certificate, a notarization key, update-signing keys and a payments store. Until then there is no download and the updater stays off.

Private beta, then launch

Not started

A beta with real Macs, thresholds tuned from what they show, the launch bar below, then 1.0. The beta channel, feedback bundle and invite flow are written.

Left: Needs the waitlist and the small web services deployed first.

After 1.0

In progress

The tools listed under "Built" are written and tested. Each ships in the release shown, once it has been checked on a real Mac.

Timeline

The original plan, in weeks. It has not been re-baselined, so treat it as an order of work and not as promised dates.

Milestones, original plan, with today's status
MilestonePlanned windowOutputStatus today
ValidateWeeks 1 to 2Landing page and waitlist, four technical spikes on real hardware, a go or no-go reviewOpen
Core engineWeeks 3 to 5Readings, sampler, database, overhead testBuilt
Detect and actWeeks 6 to 8Leak and runaway detection, safety gate, controller, rulesBuilt
Receipts, storage, app shellWeeks 9 to 10All 1.0 screens, licensing, release pipelineBuilt, except the signed release pipeline
Private betaWeeks 11 to 12Two beta builds, threshold tuningNot started
Launch 1.0Week 13Benchmark, site, launch postsNot started
1.1About 4 weeks after 1.0Startup items, Display advisorBuilt, not checked on real hardware
1.2About 6 weeks after 1.1Per-app energy, browser companionBuilt, not checked on real hardware
1.3About 6 weeks after 1.2Admin helper, power profiles, fan curvesBuilt, not checked on a signed build
macOS checkJune 2027, then each yearRe-test readings and fixes on each new macOS betaPlanned
LaterAfter 1.3Localization, a paid 2.0Planned

Built

Written and covered by automated tests, using fake Mac sources where real hardware is needed. "Built" does not mean "verified on every Mac". The tag is the release each one is planned to ship in. The code exists today, and the order can change.

Recently finished

  • Windows open no longer cost 25 to 50% of a core. With Today open it now measures 0.5% and 71 MB on one M3 Pro (decision D38).
  • Local AI headroom, power profiles, fan-curve screen and the admin helper written and tested against fakes.
  • Browser companion and its native host written. Tab titles and URLs have no field in the message format, and a test fails if one is added.
  • Per-app energy, startup items and the display advisor written and tested. Weekly report and library updates finished.
  • Independent audit fixes: Apple processes inside third-party apps are never signalled, paused apps cannot stay frozen, and a test-run guard stops tests from touching anything outside their own sandbox.
  • Overhead measured with every window closed: 0.27% of a core at the 5-second cadence, 29 MB, 0.45 wakeups a second.

Before launch

These come first. Version 1.0 is not released until they are done.

Release engineering

  • A signed and notarized build. Signed with a Developer ID with the Hardened Runtime, notarized by Apple, stapled, in a DMG, with signed updates tested from one build to the next. Until then there is no download.
  • Separate signing keys. One for the app's updates and one for the process library, so a problem with one cannot sign the other.
  • The payments setup. Stripe products and prices for the 1, 2 and 5 Mac plans, the free trial, the billing portal and refund settings, tested in Stripe's test mode first.
  • Domain and services. The domain, the waitlist and feedback services, the license and checkout service on our own server, a verified sending domain for email, and the support, privacy and security mailboxes. None of these is deployed yet, and the license service is still being built.
  • A rehearsed hotfix path, so a bad release can be replaced quickly.

Beta and launch

  • A private beta with real Macs. The plan is 50 to 100 testers, at least 30% of them on 8 GB Macs. They get weekly beta builds and an in-app way to send feedback, with a diagnostics file only if they tick it. Detection thresholds are tuned from what those Macs show, through the process library, without a new app release.
  • Benchmark runs on 8 GB and 16 GB Macs. The tests so far ran on one 18 GB M3 Pro. The public benchmark report waits for the other runs. See benchmark.
  • Launch material. Real screenshots, the benchmark page, six reviewed guides, a press kit, and a support inbox with an FAQ. The guides, the press kit and the FAQ are drafted.
  • The final prices. The prices shown today are planned prices, and Pro will be a subscription.
  • The browser extension in the browser stores. The listings go live with launch.

Checked on real Macs

Most of Hedroom is tested against fakes, which is the right way to test it but proves nothing about a real Mac. Before launch these are run on real hardware, and what they show is recorded in the project's decision log:

  • Pause and priority on real apps. Does pausing and lowering priority work on apps like Slack, Chrome's helpers, Dropbox, Docker Desktop and Spotify, from the signed build? Does a paused app resume when you click it in the Dock?
  • Readings against Activity Monitor. Per-app memory within 5% of Activity Monitor, and energy readings without admin rights.
  • Notifications, live. All eight categories and their buttons, Focus modes, and the path where notifications are turned off. The Time-Sensitive entitlement for critical alerts.
  • Calls and backups. Detection of an active call and of a running backup, because Hedroom refuses to touch those apps.
  • A first run on a clean account. Download, Gatekeeper, onboarding, permissions and the login item, on a Mac that has never seen Hedroom.
  • Storage cleanup on a real home folder. Review mode only first, then one safe category, with Put back.
  • The helper, on a signed build. The approval flow, fan control on each MacBook Pro model, and the GPU limit.
  • A live display-mode change. Including the 15-second undo and recovery if Hedroom quits during it.
  • A device matrix. An 8 GB MacBook Air M1, a 16 GB Mac, and a MacBook Pro with an external 4K display. macOS 15, 26 and 27. Light and dark, Reduce Motion, and a full VoiceOver walk of every screen.

Anything these find is fixed and the audit is re-run, with a second review of anything that touches safety.

The launch bar

Two gates, with numbers we set before seeing any data. Hedroom does not move past one until it meets them.

Beta-ready Not started

  • No known data-loss bugs, and 100% branch coverage on cleanup tests.
  • Overhead budgets met: under 0.5% of a core on average, under 80 MB.
  • Every 1.0 screen built in light and dark.
  • A notarized DMG installs on a clean Mac, and an update from the previous build works.

Launch-ready Not started

  • At least half of beta testers get a receipt with a measurable effect in their first week.
  • No more than 10% of leak notifications answered "Never touch".
  • At least 99.5% of sessions without a crash.
  • Benchmark data collected and reviewed, including 8 GB and 16 GB Macs.

After 1.0

Planned. Order and timing are not fixed.

  • 1.1, 1.2 and 1.3. Startup items and the display advisor, then per-app energy and the browser companion, then the helper, power profiles and fan curves. The code for all of them exists. Each is checked on real hardware before it ships. The helper gets a separate security review first.
  • Local AI headroom. Built, and planned to ship after 1.3.
  • A paid 2.0. Planned, at a discount for owners of 1.x. Version 1.x keeps working.
  • Languages. Hedroom is English only. Every string is already in a string catalog. German, Japanese, French and Spanish are the first candidates, and the first three languages are not chosen yet.
  • An App Store edition. A sandboxed edition with fewer features is under evaluation. A test on one Mac showed that the sandbox blocks pausing other apps, per-app energy and fan control, so it would be a smaller product, with a "Get full Hedroom" link and a purchase through the App Store. Scope, price and whether the two editions share a purchase are undecided. No work has started.
  • Optional usage statistics. If they are added, they will be opt-in and without personal data. Version 1.0 collects none, and crash reports stay local.
  • Checks against each new macOS. Every year when the betas arrive, Hedroom's readings and fixes are re-tested, starting with a check around June 2027.
  • Every release. A changelog, the update feed, and a regression run that confirms receipts still say what they should.
  • More of the process library. The signed list of plain-words explanations grows, and refreshes about once a week without a new app release.
  • Fixes from the beta. Whatever real Macs show that we did not expect goes here.

Shape the roadmap

The waitlist asks three short questions, and the answers decide what gets built first. Beta testers send feedback from inside the app. Questions or a request: contact.

Not planned

Things Hedroom is built not to be. We do not plan to add them.

  • A button that claims to free memory or make the Mac faster, or any percentage we did not measure.
  • Antivirus or malware scanning.
  • Intel Mac support.
  • Windows or other platforms.
  • A cloud account, sync, or telemetry that is on by default.
  • Safari tab sleep, because Safari extensions cannot do it.

Not in 1.0, and no commitment either way: a full app uninstaller beyond leftovers from apps you already deleted, and a duplicate-photo finder.

See also: features · requirements · trust · privacy · about · changelog

See what is slowing your Mac.

Monitoring is free, with no card and no account. Pro, with a 7-day free trial, opens with the first signed build. If it cannot show you a measurable difference, do not keep it.