Deepak (@sudo_passerby) header
Deepak (@sudo_passerby) profile picture

building voice-instinct: Instinct, but you just talk to it (open source) | 22 | SDE at a YC startup | local LLMs on Strix Halo | Arch btw :)

Bengaluru
Joined August 2023
Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)Photo by Deepak (@sudo_passerby)
Made a movie trailer for my side project lol voice-instinct: Instinct, but you just talk to it. Runs your errands in a browser on your laptop :) Foundation works, Telegram + more integrations next, still testing each step Open source, come build it: github.com/deepakkrsoni/voic…
7
3
13
1,268
:)
This is what voice-instinct actually looks like lol Asked Nova about tomorrow's reminders, cut it off mid-answer, said bye :) - STT: Parakeet (local) - TTS: Kokoro (local) - LLM: Gemini Flash, errands via Claude Code Next demo: local LLM like Gemma 12B github.com/deepakkrsoni/voic…
3
111
This is what voice-instinct actually looks like lol Asked Nova about tomorrow's reminders, cut it off mid-answer, said bye :) - STT: Parakeet (local) - TTS: Kokoro (local) - LLM: Gemini Flash, errands via Claude Code Next demo: local LLM like Gemma 12B github.com/deepakkrsoni/voic…
4
1
2
7
442
48 hours after the trailer, voice-instinct got a lot better lol - only wakes for my voice now - remembers you, files your day into markdown every night - "remind me to pay rent" actually works Linux only rn, help bring it to macOS + Windows :) github.com/deepakkrsoni/voic…
Made a movie trailer for my side project lol voice-instinct: Instinct, but you just talk to it. Runs your errands in a browser on your laptop :) Foundation works, Telegram + more integrations next, still testing each step Open source, come build it: github.com/deepakkrsoni/voic…
4
1
4
178
Btw the step screenshots are an opt-in debug mode :) Browser errands run headless, so you can't see what Nova's doing. With it on, every step gets screenshotted to your own Telegram, so you know exactly where it is in the errand and can debug when it gets stuck
2
65
Bro if it’s true then this is really crazy man, I have been using jev a lot. I’ll definitely try this out and let you know a detailed review
WAIT. An open-source decision model with 322M parameters that beats Jev on some benchmarks... Runs locally on a PHONE... And now has a FREE API?! Meet THX-01. Weights: huggingface.co/doofz/THX-01 API (Free): api.hal-x.ai/docs/thx-01 pip install thx01
1
1
152
I’m building the voice version of instinct.com
Made a movie trailer for my side project lol voice-instinct: Instinct, but you just talk to it. Runs your errands in a browser on your laptop :) Foundation works, Telegram + more integrations next, still testing each step Open source, come build it: github.com/deepakkrsoni/voic…
1
3
136
This is crazy ngl
1
2
106
This is so sick ngl
ElevenLabs founders must be SHAKING right now... these guys just dropped a voice AI you can direct like a real actor no more typing weird little codes into your script... you just write "say it like you've given up on life" and it does exactly that here's why that's a big deal: - one word comes out wrong and you change only that word, the rest of the take stays the same - it switches voices in the middle of a sentence - it voices a whole scene with several characters from one script - it clones your voice and keeps it in character in other languages and to prove it, they cloned Uncle Roger's voice and had it roast his own fried rice the whole way through lmao
1
54
also guys if you would like to contribute to voice instinct in order to make it compatible with windows and macos, that would be pretty great so that it works for windows and macos too. as i have arch linux, i made it for arch for now
2
86
My own launch triggered Nova lol It just said “Hey Nova” through my speakers and she just woke up and started listening :) And it’s using local models haha :D
1
1
3
171
Btw the voice part runs fully local on my AI Max 392: Parakeet for speech-to-text, Kokoro for the voice :) The brain is your pick: Gemini, any OpenAI-compatible API, or a local model via Ollama. Errands run through your own Claude Code
3
144
Running Qwen 3.8 27B fully on my laptop's iGPU :) 21 tok/s generation 32k context Ollama + Vulkan on Arch Not fast, but a 27B model with no dedicated GPU is kinda wild ngl
1
4
248
Yo people, what did y’all build with jev ?
1
105
Everyday in blr be like Gym -> Work -> Sleep -> Repeat :)
3
120
Finally, got it for myself :)
1
5
164
Deepak retweeted
"This is where it all starts to go wrong." 😅 Mind expanded by Architecture Patterns with Python. I always defaults to designing the DB schema first, but designing for behaviour first makes so much more sense. What's your go-to starting point when designing a new system? 🧐 👇
1
2
43
AI MAX 392 LESSGOOOO..... I've got 24 GB VRAM now :)
2
109
What's the best local LLM you guys use for coding ?
73
superpowers+ claude code >>>>
1
114