Co-founder and Head of Technology @subtlecomputing CS PhD @Stanford
Bay Area, CA
Joined January 2011
- Tweets782
- Following318
- Followers549
- Likes542
Subtle Voicebuds are Innovation Awards Honorees in the AI and headphones/personal audio categories 🥳
ces.tech/ces-innovation-awar…
Congratulations to the CES Innovation Awards 2027 honorees, whose outstanding design and engineering across 37 technology product categories stood out from nearly 4000 record-breaking submissions.
Looking forward to seeing these groundbreaking innovations shaping the future of technology at #CES2027: ces.tech/ces-innovation-awar…
1
14
314
Nothing to see here. Just 900 whisper speech input devices that can capture your voice 100 to 1000 times lower than the environment. 🤫
Get your Subtle Voicebuds here at subtle.co/voicebuds?r=ut6
—Dictated on Subtle Voicebuds
Very bad day for embedded Linux
*sshole is out
if you run sshd on alpine, openwrt or any musl box with password auth on: update now
official writeup: sshhole.com
2
1
117
Our Voicebuds debugger used to be 4+ dongles and switches for >$200, and it was very clumsy. Asked Codex to design this in 24h, and the PCB + assembly is only $30 each and all-in-one.
Still vibe coding the FW and debugging. Will post my learnings and how well it works.
You can get basically the same thing today by ordering Subtle Voicebuds 😄
Only $249, not $1.3M. 🤑
We’ve raised $1.3M from YC & Afore Capital so you never have to type or speak again.
Typing is slow. Talking into your computer is awkward. Subvocal fixes both
Our glasses give you the speed of speech without you making a sound or touching a keyboard.
It tracks your internal speech muscles with non-contact sensing and decodes silent, almost invisible mouthing into text at more than 200 words/min.
Pre-orders are now live at:
subvocal.co/preorder?utm_sou…
btw this is how it works in action: xitter.cf/jackieyang_/status/211…
This is literally the reason we built @subtlecompute Voicebuds. Being able to whisper quietly to your computer instead of speaking out loud makes a huge difference in an open office.
Would love to send you a pair to try! Demo below!
Accurate dictation anywhere, anytime.
@subtlecomputing Voicebuds
Happy Halloween!
🎃
Judging from the demos, it would work extremely well as a user simulator model for testing other voice capable systems.
Calling all AI safety researchers.
We’re inviting a small group to explore Dessa, our model trained on speech alone, and help us understand what it can and can’t do.
We’d love to hear from you: block.xyz/dessa
3
272
Attended a bunch of events at #SFTechWeek. @subtlecomputing Voicebuds won the Audience Choice Award + the VC Choice Award at Startup Grind × AI Collective’s demo night!
I’ll attend TechCrunch Disrupt next week. Happy to demo to anyone interested! Discount codes may be offered 😉
4
12
468
Exactly! This is one of the reasons we made Subtle hears only you and at a volume where other people can’t hear you.
It protects yours and others’ privacy.
As AI wearables become discreet, cheap and mainstream, they risk normalizing non-consensual recording in everyday spaces, says Joy Buolamwini
go.nature.com/4y8vaIK
I think there is a dilemma with AI agents.
Basically, when your agent is unsuccessful and unpopular, it can be very capable because nobody blocks it. But once the agent becomes capable, useful, and popular, everyone starts blocking it, and then it becomes less capable.
3
1
234
One problem with coding agents is that they don't really clean up after themselves, so the temp files/work trees of long-archived agents fill up the disk over time.
I wish that when I archive an agent, it also suggests and runs cleanup actions to remove expired work trees, etc.
2
4
321
arxiv.org/abs/2609.21967 Basically an open-weight GPT-voice-1. Very nice!
2
126
I think this is huge for audio content creation, and it's really nice that it's open-weight.
🚀 AuK is officially here. Nano banana🍌 for audio
An open-source foundation model for unified speech generation and editing.
Natural-language instructions + reference audio. One interface.
Zero-shot TTS. Instruction-controlled generation. Content editing. Whisper-conversion. De-accent. Timbre/style/emotion edit. Speed/Pitch control. Enhancement, denoising, multi-speaker and music separation.
Also releasing AuK-Flash: 4-step inference. ~4.5× faster under matched conditions.
Code, weights, and demo are live. Try it and share your feedback.
🤗 Paper & upvote: huggingface.co/papers/2609.0…
⭐ GitHub & star: github.com/Tencent-Hunyuan/A…
2
195