Makers of SKI here ā ask us anything before we launch
Hey š We're the team behind SKI, launching on Product Hunt this Thursday ā and we wanted a place to actually talk with you before then, not just pitch.
Quick what-it-is: SKI is voice coding for Claude Code, Codex and more. You talk, your agent builds, and it answers you out loud ā like a real person, not a text box. It's not dictation; it closes the loop. Everything runs on your machine (local speech in, local voice out), it's free for life, and it can even join your meetings.
We'd genuinely rather field your hard questions than sell you. Things people usually ask:
⢠How does it stay fully local ā what actually runs on-device?
⢠Which agents does it work with, and how do you connect them?
⢠The meeting side ā teammate vs. notetaker, and what's free vs. paid?
⢠The honest limits ā latency, open offices, Linux.
Ask us anything ā about the product, the decisions we made, what we'd change, whatever. We'll be around all week.


Replies
You already listed open offices as a limit, so I'll skip the obvious version and ask what's under it. There are really two loops here: voice in (I talk instead of type) and voice out (it answers me out loud). They don't fail together. Voice out survives a quiet room fine, that's just earbuds. Voice in is the half a shared room kills, and voice in is where "build at the speed you think" actually lives. So in an open office SKI seems to degrade into hearing my agent read to me while I still type, which is a pleasant TTS but not the pitch.
Which loop do you think the product actually is? If it's speaking in at thinking speed, the environments you flagged remove the whole point, not a feature, and I'd want to know what the fallback is. If it's the out-loud answer and the ambient-teammate feel, that's a different product and voice in is the optional half.
I ask because I build in a category where the moment the thing exists for is one you specifically can't make noise in, and the real work was deciding which side of the loop the product was before I could make the blocked side a first-class path instead of a sad fallback. Which side would you defend if you had to keep only one?