Update build documentation - #3

- Noted that Fleet generates a new encryption key and the original one does still remain under step #3 after a customer noted inconsistencies in our docs. 
- The statement was taken from our enforce-disk-encryption article.

- Noted that Fleet generates a new encryption key and the original one
does still remain under step #3 after a customer noted inconsistencies
in our docs.
- The statement was taken from our enforce-disk-encryption article.

…mium (#3… (#33017)

See #33006

Co-authored-by: Jacob Shandling <jacob@fleetdm.com>

…ays (#3… (#34460)

See #34420

- (#1) Swap useEffect/[showSearch] focus for native autoFocus on the
  search input. react-select unmounts <Menu> on close, so the input is a
  fresh DOM node on each open — autoFocus fires reliably without
  depending on effect deps semantics.
- (#2) Key CustomMenuList's useLayoutEffect on React.Children.count(...)
  so it only re-measures scroll dimensions when the option list actually
  changes (search filter shrink/expand). onScroll still covers user-
  driven position changes.
- (#3) menu.zIndex 6 -> 99 (page-overlay tier), __scroll-fade z-index
  1 -> 9 (in-flow overlay tier), matching the 9/99/999 convention.
- (#4) Add optional onAddFleet callback prop. If provided, fires
  instead of the browserHistory.push fallback — lets consumers use
  their own router.push. Nine current callers unaffected.
- (#8) TODOL -> TODO: typo in the tests file (inherited from the
  deleted TeamsDropdown.tests.tsx).

  stale comment, side effects in state updater

- (#2) Rework the scroll-fade to a 0-height sticky anchor with an
  absolutely-positioned ::before pseudo-element toggled via opacity on
  a --visible modifier. The old approach mounted/unmounted a 35px
  sticky div, so MenuList.scrollHeight grew and shrank as the fade
  toggled — browsers can clamp scrollTop when scrollHeight shrinks,
  causing a visible jump near the end of the list. The anchor is
  always in the DOM at 0 flow height; scroll metrics stay stable.

- (#3) Chain react-select's own innerProps.onScroll and innerProps.
  onMouseDown before our custom logic. The previous spread-then-
  override pattern silently dropped whatever react-select provided,
  which could break its own scroll-to-highlighted-option or focus
  tracking on a future version.

- (#5) Fix stale comment on the hidden-input-focus useEffect. It said
  "CustomMenuList focuses the search input on mount," but the search
  input now lives in CustomMenu and uses native autoFocus. Corrected
  the comment to reflect where focus actually comes from.

- (#6) Move side effects (onOpen? / setSearchQuery) out of the
  setMenuIsOpen state updater in toggleMenu. React's Strict Mode
  double-invokes state updaters, which would double-fire onOpen and
  double-clear searchQuery. Reading menuIsOpen directly is safe — a
  single user click can't race with itself.

- (#3) Deleted orphaned components/UploadListHeading/ folder — no
  external importers left after the tab-header refactor.
- (#4) ConfigurationProfiles: hide the __tab-header (description +
  Add button) when !mdmEnabled so the "Create and upload configuration
  profiles..." copy no longer contradicts the "MDM must be turned on"
  EmptyState directly below.
- (#5) Extracted the duplicated __tab-header block into a
  @mixin tab-header in styles/var/mixins.scss; six copies across five
  files now @include it.
- (#6) Comment on .profiles-tab explaining why ConfigurationProfiles
  splits its card wrapper from its per-tab layout (two inner tabs
  mirroring .assets-tab) while Certificates / ScriptLibrary stay
  single-block (one-tab card).
- (#7) Comment on <span className="controls"> in TableContainer.tsx
  flagging the load-bearing shape that _styles.scss's :has(.controls
  > *) collapse rule depends on.

…our inventory doesn't have (#52818)

**Related issue:** NA

New article drafted from an approved content pitch in
#_content-news-digest, covering Microsoft's Project Zenith (a
ready-to-code Windows 11 setup for local AI development on AMD's Ryzen
AI Halo hardware) and why these machines are still endpoints that need
to show up in inventory.

# Checklist for submitter

- [ ] Changes file added for user-visible changes in `changes/`,
`orbit/changes/` or `ee/fleetd-chrome/changes`.
See [Changes
files](https://github.com/fleetdm/fleet/blob/main/docs/Contributing/guides/committing-changes.md#changes-files)
for more information.

## Testing

- [x] QA'd all new/changed functionality manually (reviewed against
`fleet-article-formatting` and `content-style` skills; verified links,
category value, and description length)

### Title selection

Checked the last 10 merged/open PRs from this pipeline: 7 of 10 use the
"What X means for Y" template, and the last 3 in a row all use it. To
break that pattern, I brainstormed 5 candidates with distinct
structures:

1. "Project Zenith needs 64 GB of memory, and Fleet needs to see it"
(parallel two-clause)
2. "Windows machines running 30-billion-parameter AI models are already
on your network" (plain declarative)
3. **"AMD's Ryzen AI Halo chips created an endpoint category your
inventory doesn't have"** (declarative claim naming the stakes) —
selected
4. "Microsoft built a Windows 11 mode for local AI models over 30B
parameters, and it's already unmanaged" (compound cause/effect)
5. "64GB of unified memory, one Ryzen AI Halo chip, zero visibility"
(list/triad fragment)

Picked #3 over runner-up #5: it names the actual news hook (the hardware
spec) while stating the reader's problem plainly ("an endpoint category
your inventory doesn't have"), which reads clearer to an IT/security
audience skimming a feed than #5's punchier but more cryptic fragment.

## AI

**AI:** Claude Code (claude-sonnet-5)

https://claude.ai/code/session_01MgZRQpzyznZLfXuqc4he3Q

---
_Generated by [Claude
Code](https://claude.ai/code/session_01MgZRQpzyznZLfXuqc4he3Q)_

---------

Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: robinedev <aubedev@outlook.com>