Explore moving modifiers to the thumb cluster

[rdlu]

Goal

Relocate the home-row-mod responsibility from bottom-row letters (current state on main after #6) to the thumb cluster, while still preserving:

• easy hold access to all four non-base layers (NAV, CODE, MEDIA, SYS|NUM)
• caps_word key
• fast access to DEL
• BSPC, SPACE, ENTER on the thumbs (current behavior)
• all four mods directly holdable as real chords — the workflow needs arbitrary combinations of CTRL + ALT + GUI + SHIFT + key (many apps use 3-mod shortcuts)

Non-constraint: games. Game sessions will run on a dedicated locked layer (see "Future: game layer" below), so the daily layout does not need to preserve held-SPACE / held-WASD semantics. This unlocks lt NAV SPACE and similar hold-on-common-key patterns.

No rush — this is a design exploration issue, not a queued task.

Current thumb cluster (row 4, positions 52–63)

Left  (52–57): [C_MUTE enc] [mo SYS] [F11] [LGUI] [mo NAV] [BSPC]
Right (58–63): [ENTER] [SPACE] [lt CODE DEL] [mo MEDIA] [F12] [caps_word]

5 usable left + 6 usable right = 11 thumb slots (excluding encoder click).

Demand list

Need	Count
Layer holds (NAV, CODE, MEDIA, SYS)	4
Modifiers (LCTRL, LSHFT, LALT, LGUI) — all must be directly holdable	4
Text keys (SPACE, ENTER, BSPC, DEL)	4
caps_word	1
Nice-to-have (F11, F12, C_MUTE encoder)	3

16 demands, 11 slots → we must combine via hold-taps. Layer-based mods are off the table because any multi-mod chord needs all mods holdable at once.

---

Idea 1 — Mod-tap thumbs, plain mo for layers (urob-style)

Every text-carrying thumb key becomes a mod-tap. Layers remain plain mo / lt. require-prior-idle-ms = 150 protects the tap side when chording (press mod, then tap the paired text key within the idle window → the text key resolves to tap, not hold).

Chord compatibility — the hard constraints

Assume pairings: LCTRL=BSPC, LSHFT=SPACE or ENTER, LGUI=F11, LALT=?

Shortcut	Needs	Reachable?	Notes
CTRL+P	LCTRL + P	✅	left thumb + right hand
CTRL+SHIFT+P	LCTRL + LSHFT + P	✅	two left-thumb keys (BSPC + SPACE) + right P — thumb roll
CTRL+C / CTRL+V	LCTRL + letter	✅	clean
CTRL+TAB	LCTRL + TAB	✅	TAB is on top row, not thumb
ALT+TAB	LALT + TAB	✅	as long as LALT is not paired with TAB
ALT+F4	LALT + F4	✅	F4 on NAV layer — NAV held + ALT thumb + F4 key
GUI+L	LGUI + L	✅	clean
GUI+SHIFT+S	LGUI + LSHFT + S	⚠️	if both on left thumb and S is left-hand, awkward. Distribute mods across both thumbs
CTRL+ALT+DEL	LCTRL + LALT + DEL	❌ if mt LALT DEL	holding DEL produces ALT, not DEL. Fix: don't pair LALT with DEL. Keep DEL on lt CODE DEL so tap-for-DEL stays clean
CTRL+SPACE (autocomplete)	LCTRL + SPACE	✅	LCTRL held, SPACE pressed within idle window → SPACE stays as tap
CTRL+ALT+SHIFT+letter	all three + letter	⚠️	three thumb keys held simultaneously — possible but clumsy; needs mods distributed across both thumbs

Key design decisions for Idea 1

1. Which text key pairs with LALT? Must not be commonly chorded with ALT. Candidates:
   • F11 or F12 (rare enough) — but then those thumb slots become the ALT home
   • ENTER (ALT+ENTER used by some apps for fullscreen / properties — moderate risk)
   • A dedicated plain-LALT slot, no tap side (costs a slot)
2. Bilateral SHIFT — strong case for putting LSHFT on both thumbs: left for right-hand capitals, right for left-hand capitals. Costs an extra slot.
3. F11 / F12 / LGUI — demote to SYS or NAV layer to free thumb slots for mods.
4. Caps word — outer-edge slot (pos 63) stays.

Three concrete variants to build

Variant A — bilateral shift, F11/F12 carry GUI/ALT

Left:  [mo SYS] [mo NAV] [mt LSHFT SPACE] [mt LCTRL BSPC] [mt LALT F11]
Right: [mt LSHFT ENTER] [lt CODE DEL] [mo MEDIA] [mt LGUI F12] [caps_word]

• SPACE moves to left thumb (swap with current); gain bilateral shift
• LALT on F11 (rare tap side, acceptable)
• LGUI on F12 (ditto)
• Awkward chord: CTRL+ALT+DEL — CTRL (BSPC) + ALT (F11) both on left thumb, DEL on right. Doable via thumb roll.
• Awkward chord: GUI+SHIFT+S — GUI on right, SHIFT on right (ENTER), S is left-hand. Two right-thumb keys + left hand → clean if F12 and ENTER are reachable by thumb roll.

Variant B — layer-taps for layers, mods spread across both thumbs

Left:  [mo SYS] [mt LGUI F11] [mt LCTRL BSPC] [lt NAV SPACE] [mt LSHFT ENTER-dup]
Right: [mt LSHFT SPACE-dup?] [mt LALT DEL?] [lt CODE F12?] [mo MEDIA] [caps_word]

• Combines layer-tap on SPACE (now that games aren't a constraint)
• Dense — may overflow 11 slots
• Similar awkward edges

Variant C — demote F11/F12/LGUI to SYS layer, open thumb slots for dedicated mods

Left:  [mo SYS] [LALT-plain] [mt LCTRL BSPC] [mt LSHFT SPACE] [mo NAV]
Right: [mt LSHFT ENTER] [mt LGUI DEL-or-other] [lt CODE DEL] [mo MEDIA] [caps_word] [—]

• Plain LALT thumb slot (no tap side) — costs one slot but eliminates the LALT pairing problem
• F11/F12 accessible on SYS layer
• Probably the cleanest of the three, at the cost of losing direct thumb-F11/F12

None of A/B/C are obviously right. Picking means deciding which chord we're willing to make slightly awkward and how precious F11/F12 are as thumb slots.

---

Idea 4 — Home-row mods back on home row, NAV arrows relocated instead

Alternative framing: instead of moving mods onto thumbs, solve the NAV arrow repeat problem by moving the arrow cluster off positions with hold-taps underneath.

Options for new NAV arrow location:

• NAV layer row 3 (base N/M/,/. — all plain kp currently) — shift vim-style arrow cluster down one row
• Dedicated physical arrow keys on base (pos 32, 45) — make them &trans on NAV so base arrows pass through
• Thumb-based arrows (mo NAV + thumb keys for L/R, separate U/D binding)

Pros

• Home-row-mod ergonomics restored (the most thoroughly-documented HRM pattern, decade of tuning)
• No thumb hold-taps → thumbs stay crisp
• All four mods holdable on home row — chords are clean
• Bottom row also stays plain

Cons

• Loses hjkl-on-home-row muscle memory on NAV
• Doesn't fit the original "mods on thumbs" premise of this issue

---

Future: dedicated game layer(s)

Tracked separately from this design work — note it here so the daily layout doesn't try to preserve gaming semantics.

Sketch:

• New top-level layer (e.g. GAME), entered via &to GAME from the SYS layer (explicit toggle, not momentary)
• Exit via an easy-to-reach key on the game layer itself (e.g. long-press ESC → to BASE)
• Game layer disables mod-taps entirely — pure plain &kp for WASD + held-SPACE / held-SHIFT semantics
• Possibly multiple game layers (per-game profiles) toggled from SYS

Open design questions for the game layer:

• One layer or per-game variants?
• How to indicate "game mode active" — RGB color? OLED display?
• Should the game layer keep NAV access for non-gaming keys (alt-tab out, discord messages), or be a pure island?
• Bluetooth profile switching still accessible?

Not in scope for this issue — file a separate issue when ready to design.

---

Open questions (daily layout)

• Are F11 / F12 / LGUI precious enough to keep dedicated thumb slots, or demote to a layer?
• Which single multi-mod shortcut hurts the most if it becomes a two-step gesture?
• Preference: bilateral SHIFT (extra slot) vs. compact single-side SHIFT?

Suggested next step

Given the chord-compatibility analysis, Idea 4 is looking stronger than Idea 1 for a workflow heavy on multi-mod shortcuts. Home-row mods have a decade of refinement behind them; thumb-cluster mods are harder to tune and each variant trades one awkward chord for another.

Options:

1. Build Idea 4 as a canary: keep Move home-row mods to bottom row to unblock NAV arrow repeat #6's working layout as the safety net on main, branch off to restore home-row mods + move NAV arrows elsewhere. Probably the path of least regret.
2. Build Idea 1 Variant C as a canary: plain LALT thumb + dedicated mod slots, demote F11/F12 to SYS. Cleanest of the thumb-mod variants.
3. Park this issue until there's a concrete pain point with Move home-row mods to bottom row to unblock NAV arrow repeat #6's bottom-row mods.

Related: #5 (original NAV arrow repeat), #6 (bottom-row-mods fallback on branch feat/nav-arrow-repeat).

[rdlu]

Three draft PRs opened for side-by-side SVG comparison

All three are draft design explorations — they branch off main and regenerate the keymap SVG so the rendered layout can be compared visually. None are meant to merge; close whichever don't win.

• [design] Variant 1A — bilateral shift, thumb mod-taps #8 — Variant 1A (bilateral shift, thumb mod-taps, F11/F12 carry GUI/ALT)
• [design] Variant 1C — plain mod thumbs #9 — Variant 1C (plain mod thumbs, lt NAV SPACE)
• Relocate NAV arrows + move LALT to S/L and RALT to R/U #10 — Idea 4 (keep home-row mods, relocate NAV arrows one row south)

Rendered SVGs are committed on each branch at keymap-drawer/eyelash_sofle.svg — open the "Files changed" tab on each PR or browse the file on the branch.

My read after drawing them: Idea 4 (#10) is still the lowest-regret path given the multi-mod chord requirement. Variant 1C (#9) is the cleanest thumb-mod variant but loses bilateral shift. Variant 1A (#8) preserves bilateral shift but the CTRL+ALT+DEL thumb geometry is awkward.