Developers Jason and Nick are trading notes on how rigid slash commands clash with real-world hands-free dictation habits, where one prefers fluid voice phrases while pacing and the other flags how even small voice-to-text slips derail custom skill triggers.
The issue with having skills and subcommands like `/loop` or `/goal` Is that, fundamentally, I have to be close to a keyboard?
Whereas most of my actual work is through dictation. I'm walking around my house, pacing back and forth, and I've really learned to just start using things like
"Keep an eye on this" "Just go ahead and set a goal"
I would never type it, but saying it is just as fast.
Nick Dobos notes that voice dictation frequently mangles the exact names chosen for skills or subcommands, breaking the matching logic in tools that rely on those precise triggers during hands-free sessions.
Jason points out that keyboard-dependent commands like /loop or /goal slow down work compared with spoken instructions, though the discussion leaves open whether current dictation tools can reliably support the needed precision.
Many users are excited about voice commands with AI tools like Codex as a huge productivity unlock enabling autonomous goal-setting, while some criticize current UX blindspots such as poor keyboard integration and voice typos.
No Digg Deeper questions have been answered for this story yet.
@jxnlco The absolute worst when you make a skill with a name you love only to discover that voice mode consistently typo’s it
The issue with having skills and subcommands like `/loop` or `/goal` Is that, fundamentally, I have to be close to a keyboard?
Whereas most of my actual work is through dictation. I'm walking around my house, pacing back and forth, and I've really learned to just start using things like
"Keep an eye on this" "Just go ahead and set a goal"
I would never type it, but saying it is just as fast.
I've also just added the following in my home directory AGENTS[.]md
[appshot this, and ask Codex to do that too]
""" # Goal Setting
- Before starting substantive work on any task that appears moderately difficult or likely to take roughly 10 minutes or more, always create an active goal with `create_goal`, even if the user did not explicitly ask for one.
- Write the goal as a concise execution contract with a verifiable completion criterion. Keep one coherent objective per goal and preserve the user's constraints, named artifacts, and required validation.
- For difficult, exploratory, or evolving tasks, revisit and dynamically refine the active goal as understanding, constraints, or the evidence needed for completion changes. Preserve the user's intent, and do not silently weaken the completion criterion or broaden the scope.
- Keep the goal active through implementation and verification. When the objective is genuinely achieved and no required work remains, mark it complete with `update_goal`; use blocked status only under the tool's stated blocked criteria.
- Do not create goals for trivial requests that can be answered or completed immediately in a single short step. """
@jxnlco this is the way
@jxnlco How do you initialalise plugins and skills when dictating?
@jxnlco you might like this, it's not very good yet but lets codex loop in the background while you talk to a voice agent
@BasedDaedalus I use a mix of this and Wispr Flow, to be honest.
@jxnlco is codex's native dictation tool good enough for this? idk if i'm a text purist, but i think a well-written prompt works way better than just talking out loud.
@jxnlco Advanced voice mode + Codex would be AGI.
@jscraik "use slack"
@jxnlco i guess i'll have to give it a try. do you sometimes edit the text before sending it, or do you usually just leave it as is?
@AnatoliDev I'll keep that in mind.
@jxnlco it should just be smart enough to use skills like goal and loop on its own based on what you're asking to the extent where we don't have to directly type /[skill]
@BasedDaedalus I have something in my prompt that says that most of my inputs are from dictation. If there are confusing words, try to spell it out phonetically and see if you could figure out what the typo was.
@jxnlco Interesting I’m gonna give this a try today
@jscraik @jxnlco When I tell it to use test and deploy it fires my /test-and-deploy skill?
@jxnlco telling codex to set its own goal instead of manually typing /goal has been a huge unlock for me lately. Asking my Chief of Threads to set itself a goal to deliver a plan by spinning up task-specific threads has totally changed the way I develop.
@jxnlco So you dictate to slack mention skills and plugins and then codex app picks that up, this would be a good workflow blog as I have been trying to figure this out for my brain injury👊🏻🇬🇧
@jxnlco this is such an underrated UX blindspot.
voice-first tools and keyboard-first design still dont talk to each other properly.
@jxnlco The impact on coworking cafes due to dictation led engineering is going to be real 😂
@jxnlco You only need slash commands if you haven't automated properly