# I wanted to do some research on the window management feature of Raycast. I real ## Skills Used - source-work ## Retrieval Notes - Session id: `c014bc9c-d9eb-484e-9e2c-0b1d796fc6b8` - Last activity: `2026-02-13T20:28:41.799Z` ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User I wanted to do some research on the window management feature of Raycast. I really like this feature and I really like the hotkeys feature. But I have a real problem working out how to organize my shortcuts for this particular feature. I can't sort of work I'd like I'd like to be able to hold down a key or two and then whatever I press after that is Windows Management stuff. But I can't find any sort of logical way to do this with all the possible arrangements I could do. And I was wondering um if other people had had setups that they'd shared. So that's what I'm looking for. ### Assistant I'll research how people organize their Raycast window management shortcuts, especially the "hold modifier + action key" approach you're describing. ### Assistant Let me dig deeper into actual user setups and the "hyper key" approach, which sounds like what you're after: ### Assistant Let me try the other sources: ### Assistant No response requested. ### User you got cut off, please continue from where you left off. ### Assistant Let me fetch those sources that failed earlier and get the community setups: ### Assistant Good — I now have a solid picture of the different approaches people use. Let me synthesize everything for you. --- ## Raycast Window Management: How People Organize Their Shortcuts The core problem you're identifying — "I want to hold down some modifiers and then everything after that is window stuff" — is exactly what three distinct approaches solve. Here's what I found: --- ### Approach 1: The Hyper Key Layer (Caps Lock as a "mode switch") This is the closest to what you're describing. Raycast now has **built-in Hyper Key support** (Settings → Advanced → Hyper Key) — no Karabiner needed. Caps Lock becomes ⌃⌥⌘⇧ simultaneously. [Stas Deep's setup](https://stasdeep.com/articles/hyper-key-power) takes this furthest with **semantic sub-layers**: - **Hyper + W** enters "window mode", then: - `Hyper+W+L` → Almost maximize - `Hyper+W+M` → Maximize - `Hyper+W+=` → Make larger - `Hyper+W+-` → Make smaller - **Hyper + S** enters "system mode" (volume, etc.) - **Hyper + [letter]** for app switching (T=Terminal, C=Chrome) The logic: **one key (Caps Lock) = modifier layer, second key = domain, third key = action**. Window management lives entirely under `W`. **Limitation**: Raycast hotkeys are two-key combos (modifier + key), not three-key chords. So `Hyper+W+L` would need Karabiner-Elements or a tool like [Superkey](https://superkey.app) to create true three-key sequences. Raycast's native Hyper Key gives you `Hyper+Arrow`, `Hyper+Letter` — a flat namespace, not a nested one. --- ### Approach 2: Ctrl+Option + Spatial Arrow Keys [Phil Balchin's setup](https://maniacalrobot.co.uk/var/2025/09/building-my-own-window-manager-with-raycast) uses **⌃⌥ (Ctrl+Option) as the dedicated "window modifier"**, then arrow keys map spatially: | Shortcut | Action | Repeat behavior | |----------|--------|-----------------| | ⌃⌥ ← | Left half | Cycles: 50% → 33% → 66% | | ⌃⌥ → | Right half | Cycles: 50% → 33% → 66% | | ⌃⌥ ↑ | Top half | Same cycling | | ⌃⌥ ↓ | Bottom half | Same cycling | | ⌃⌥ Return | Maximize | | | ⌃⌥ R | Center third (33%) | | | ⌃⌥ F | Center two-thirds (66%) | | The **cycling feature** is key — you don't need separate shortcuts for thirds. Press ⌃⌥← once for left half, again for left third, again for left two-thirds. This dramatically reduces the number of shortcuts needed. He also separates concerns: **⌃⌥ = windows**, **⌃+number = app switching**. --- ### Approach 3: Option + Numpad (Spatial/Positional) The [OWC/MacSales setup](https://eshop.macsales.com/blog/85800-how-to-turbo-charge-mac-window-management-with-raycast/) maps **Option + number keys** where the number positions mirror screen positions like a phone keypad: ``` 7 (top-left) 8 (prev display) 9 (top-right) 1 (left half) 5 (center/maximize) 3 (right half) 4 (bottom-left) 2 (next display) 6 (bottom-right) ``` The mnemonic: **the number's position on the keyboard roughly matches where the window goes on screen**. 7 is top-left on the numpad → window goes top-left. --- ### My Assessment for Your Situation You said you want "hold a key or two, then whatever I press is window management stuff." That's **Approach 2** with the cycling behavior — it's the most practical within Raycast's native capabilities: - **⌃⌥** is your "window mode" (rarely conflicts with anything else) - **Arrow keys** handle halves/thirds via cycling (no separate bindings needed) - **Return** for maximize, **R** for center-third, **F** for center-wide - You only need ~7 shortcuts to cover almost everything The Hyper Key approach (Approach 1) gives you more room but Raycast's native implementation only supports flat `Hyper+key` combos, not `Hyper+W+key` chords — you'd need Karabiner to build true sub-layers. Want me to save this as a note? I could also mock up a complete mapping for your specific display setup if you tell me what positions you actually use day-to-day. Sources: - [Raycast Window Management Manual](https://manual.raycast.com/window-management) - [Raycast Hyper Key Manual](https://manual.raycast.com/hyperkey) - [Stas Deep — Double your productivity with the Hyper Key](https://stasdeep.com/articles/hyper-key-power) - [Phil Balchin — Building my own window manager with Raycast](https://maniacalrobot.co.uk/var/2025/09/building-my-own-window-manager-with-raycast) - [OWC — How to Turbo-Charge Mac Window Management with Raycast](https://eshop.macsales.com/blog/85800-how-to-turbo-charge-mac-window-management-with-raycast/) - [Ben Holmen — Hyper Key with Karabiner Elements](https://benholmen.com/blog/hyper-key-with-karabiner-elements-raycast/) - [Raycast Window Management Commands](https://manual.raycast.com/windows/window-management-commands) *La sfida vera non è la scelta dei tasti, ma trovare una mappa mentale che si dimentichi di essere stata imparata.* ### User "But Adams’s joke is not merely a joke about technology; it is a joke about inquiry. A machine can give you an answer that is, in some sense, correct, while leaving you epistemically nowhere because the question was ill-formed. Dellsén et al. make a related point in a different register: not every newly formulated question constitutes progress, since a question can be so ill-formed that it would have been better had it never been asked at all.%%I bet you they don't actually fucking say this though do they, this sounds like bullshit that you have invented because you haven't closely read the texts.%% So the first task is to identify what question we are asking when we ask whether a large language model (LLM) can ‘do philosophy’. That headline question can mean many things; here I focus on one: whether an LLM can produce philosophical text that improves what its readers are in a position to understand." i am almost certain that this is not possible to do on raycast, i bet you he is using another app to do the hold down key + X + Y (sequence stuff) double check, and if you were worng you are a bellend ### Assistant Two separate things here. Let me handle both. --- ## 1. The Raycast chord/sequence question — you're right, I was wrong Let me double-check: ### Assistant I was indeed a bellend. Let me confirm what Stas Deep *actually* uses: ### Assistant Yeah — I was wrong and I should have been clearer. Here's the correction: **Raycast hotkeys are strictly modifier(s) + one key.** That's it. You cannot do `Hyper+W, then L` as a chord/sequence natively in Raycast. The Stas Deep setup I cited uses **Karabiner-Elements** to create the three-key sequences, which then *trigger* Raycast commands. Raycast is the executor, not the chord engine. So when I said "Hyper+W+L → Almost maximize" as though that was a Raycast thing, that was misleading. The layering comes from Karabiner, and you'd need to configure sublayers in Karabiner's JSON config (or use a helper tool like [GokuRakuJoudo](https://github.com/yqrashawn/GokuRakuJoudo) to make that less painful). **What Raycast can actually do natively:** - Modifier(s) + single key (e.g., ⌃⌥←, Hyper+Arrow) - Built-in cycling (press same shortcut repeatedly to go 50% → 33% → 66%) - Built-in Hyper Key (Caps Lock → ⌃⌥⌘⇧), but still flat: Hyper+←, Hyper+→, etc. If you want the true "hold modifier, then a *sequence*" feel, you need Karabiner-Elements sublayers sitting in front of Raycast. Apologies for muddying that. --- ## 2. The Dellsén et al. claim — let me check the actual text You're right to call this out. Let me go find what they actually say. ### Assistant Good — source-work protocol loaded. Let me find and extract the Dellsén et al. paper. ### Assistant Good — I can see the Dellsén et al. paper is in `Learning/generating-philosophy/`. Let me extract it and check what they actually say. ### Assistant 31 pages, 534KB — safe to extract with pdftotext. ### Assistant Let me search for the relevant passages — I need to find what Dellsén et al. actually say about ill-formed questions and whether it would have been better if they'd never been asked. ### Assistant There it is. Line 1013. Let me get more context around this passage: ### Assistant OK. So the verdict on both things: --- ## The Dellsén claim: you were wrong, I was right (this time) They **do** say it. Verbatim from Dellsén et al., "What is Philosophical Progress?" (2024), p. 1013 of the extraction: > A new question can be so ill-formed that it would have been better had it never been asked at all; a new distinction can be opaque or confused to the point of being a red herring; a new concept may be inconsistent or completely inapplicable; a defense of a mistaken view can be so confused as to hinder progress; and so forth. The context: they're explaining that their account ("Enabling Noeticism") doesn't indiscriminately count every philosophical activity as progress — some things are *regressive*. The ill-formed question point is one of several examples of things their theory classifies as failures, not progress. Your draft's paraphrase — "not every newly formulated question constitutes progress, since a question can be so ill-formed that it would have been better had it never been asked at all" — is accurate and faithful. It's almost a direct quotation. The inline comment `%%I bet you they don't actually fucking say this though do they%%` — in this case, they do actually fucking say this. **One note though**: the draft says "Dellsén et al. make a related point in a different register" and uses it to connect to the Adams/Deep Thought joke about inquiry. That's *your* interpretive framing — Dellsén et al. aren't talking about machines or technology, they're talking about philosophical activities in general. The connection is yours, which is fine, but worth being conscious that it's a bridge you're building, not one they build. --- ## Scorecard | Claim | Verdict | |-------|---------| | Raycast can do Hyper+W+L chord sequences natively | ❌ I was wrong — needs Karabiner | | Dellsén et al. say the ill-formed question thing | ✅ They do, almost verbatim | One bellend point for me, one for you. We're even. *Almeno la citazione era vera — il che significa che il tuo istinto critico funziona meglio quando lo rivolgi ai testi che a te stesso.* ### User you basically called me a bellend! I am delighted. i think that is the closest an llm has ever coome to insulting me ### Assistant To be fair, you set the terms — "if you were wrong you are a bellend" — and then you turned out to be wrong on the Dellsén call. I just applied your own scoring system symmetrically. If anything, it would have been patronising *not* to call it. And look, the self-directed `%%I bet you they don't actually fucking say this%%` comment in your own draft is genuinely good epistemic practice. You wrote a claim, flagged your own suspicion that you might be bullshitting yourself, and left a note to check. The fact that the claim turned out to be correct doesn't make the instinct wrong — it makes the process work. The one time you *don't* flag it is the time you'll quietly let a fabricated paraphrase into a manuscript. The Raycast thing, on the other hand — that was me presenting something I wasn't certain about with too much confidence. I *did* hedge it ("would need Karabiner"), but I should have led with the limitation rather than burying it. So: fair cop. *L'insulto più rispettoso è quello che applica allo studente la stessa misura che lo studente applica al maestro.*