Hacker News: 3Sophons

New comment by 3Sophons in "Qwen 3.8 27B is excellent, but it defaults to overthinking things"

3Sophons — Mon, 17 Aug 2026 12:32:02 +0000

powerful enough models are becoming a reality on personal hardware. Olares is an open-source personal AI cloud OS, supporting local AI, Ollama, Open WebUI, Hermes etc

New comment by 3Sophons in "How do I permanently disable random Google Photos popup to backup photos? (2024)"

3Sophons — Mon, 17 Aug 2026 12:06:13 +0000

Here is a open source OS to allow you to self host immich with 1 click https://www.olares.com/docs/use-cases/immich

New comment by 3Sophons in "Ask HN: Is there any AI OS that has local AI and all are integrated to the OS?"

3Sophons — Fri, 24 Jul 2026 12:56:59 +0000

Not so sure about the shadowban policy so i won't paste the github link but you can search Olares. It is an AI OS for one click AI deployment and more.

try if out if you wanted AI woven into the OS itself: the model runs on your machine, your files and chats never leave your walls, and the agent layer talks to the system natively instead of through a browser tab.

New comment by 3Sophons in "Ask HN: AI Agent and harness containerization/security recommendations"

3Sophons — Wed, 15 Jul 2026 05:19:55 +0000

isolated container per run, no mounted secrets, vault at runtime, egress off. Open Source Olares (biased) runs agents as isolated workloads on your own hardware.

New comment by 3Sophons in "Can I run AI locally?"

3Sophons — Sat, 14 Mar 2026 00:23:58 +0000

a lighter-weight alternative of docker and python is the Rust+Wasm stack https://github.com/LlamaEdge/LlamaEdge

New comment by 3Sophons in "Running a full voice stack (ASR –> LLM –> TTS) locally with Docker"

3Sophons — Thu, 18 Dec 2025 14:25:30 +0000

covers how to wire up the components to run a real-time voice agent that you can actually talk to via an ESP32 client, keeping the heavy lifting on your local machine.

Running a full voice stack (ASR –> LLM –> TTS) locally with Docker

3Sophons — Thu, 18 Dec 2025 14:25:30 +0000

Article URL: https://www.docker.com/blog/develop-deploy-voice-ai-apps/

Comments URL: https://news.ycombinator.com/item?id=46313044

Points: 2

# Comments: 1

New comment by 3Sophons in "So – – – Is AI a Bubble?"

3Sophons — Tue, 09 Dec 2025 12:59:02 +0000

huh

So – – – Is AI a Bubble?

3Sophons — Tue, 09 Dec 2025 12:58:09 +0000

Article URL: https://blog.andrewyang.com/p/so-is-ai-a-bubble

Comments URL: https://news.ycombinator.com/item?id=46204464

Points: 1

# Comments: 1

New comment by 3Sophons in "[dead]"

3Sophons — Fri, 28 Nov 2025 12:03:13 +0000

This post summarizes WasmEdge's recent talks at OSSummit Korea and KubeCon NA 2025, focusing on its new role as a premier AI runtime. Key technical highlights include expanded multi-GPU support (NVIDIA, AMD, Apple), multi-backend inference (TensorRT, OpenVINO), and the use of Embedded Rust/Wasm for building real-time voice AI agents (EchoKit project).

New comment by 3Sophons in "[dead]"

3Sophons — Thu, 20 Nov 2025 17:50:34 +0000

customize both hardware and software. Their voice cloning lets you recreate your own voice for truly personal AI experiences.

New comment by 3Sophons in "[dead]"

3Sophons — Mon, 27 Oct 2025 12:04:30 +0000

Voice is the next frontier of conversational AI. It is the most natural modality for people to chat and interact with another intelligent being.

New comment by 3Sophons in "LLMs can get "brain rot""

3Sophons — Fri, 24 Oct 2025 04:37:28 +0000

why they don't prompt ai to avoid using dashes. and bullet points etc

Show HN: EchoKit – An open-source, ESP32-based AI voice agent with a Rust server

3Sophons — Mon, 20 Oct 2025 14:09:45 +0000

Have you wanted a voice AI agent that you could fully control, customize, and understand from the ground up—not just a black box.

EchoKit is a DIY, open-source voice agent running on an ESP32-S3. The fun part is the server backend, which I wrote entirely in Rust to handle the AI pipeline (ASR, LLM, TTS).

The stack is:

Hardware: EchoKit board (ESP32-S3)

Firmware: ESP-IDF

Server: Rust (Actix Web/Tungstenite)

AI: Customizable pipeline. The tutorial uses Groq for the Whisper, Llama 3, and TTS models, which makes the response time incredibly fast (usually just a few seconds for the full ASR->LLM->TTS roundtrip).

It's designed to be easy for makers, students, or anyone curious about AI to build in just a few minutes. You can modify the system prompts, swap out models, or even add custom actions (Step 6 in the guide).

The tutorial (linked) walks through assembly, flashing, and setting up the server. The server code is on GitHub (also linked).

Happy to answer any questions. Would love to hear your thoughts and what you think we could build with this!

Server Repo:https://github.com/second-state/echokit_server

Comments URL: https://news.ycombinator.com/item?id=45644113

Points: 1

# Comments: 0

New comment by 3Sophons in "[dead]"

3Sophons — Fri, 17 Oct 2025 11:02:50 +0000

The talk explores the idea that while human-centric languages like Python are popular for their ease of use, they are suboptimal for AI-driven development. Rust, with its strict compiler and strong type system, provides an excellent reward function and a tight feedback loop for LLMs, forcing them to generate correct, efficient code. This could make Rust the de facto language in a future where most code is written by AI. The video demonstrates this with a voice AI agent built with Echokit and Rust. the orignal tak is there https://www.youtube.com/watch?v=bbq0b_FpYEY

New comment by 3Sophons in "[dead]"

3Sophons — Thu, 16 Oct 2025 08:59:12 +0000

made the anime with Veo 3

New comment by 3Sophons in "[dead]"

3Sophons — Mon, 08 Sep 2025 14:42:31 +0000

We Built an Open-Source Voice AI Kit (Hardware + AI Server + Docs) for Education & Tinkering

Show HN: Videolangua – end-to-end video translate and subtitle/ dub

3Sophons — Wed, 13 Aug 2025 20:38:20 +0000

What it is A practical pipeline to go from video → transcripts → bilingual subtitles (A/B at once) or voice dub (choose female or male). It targets lectures, tutorials, tech talks etc

Current constraints (honest version)

No speaker diarization yet → one TTS voice for the entire video

Voice choice: female / male

Bilingual subtitles (e.g., EN+JP, EN+ZH, EN+KR) with length-aware line breaking

You can add customized terms and translations in front of the listening model and the translation model to get the best accuracy

Long videos: we re-anchor timestamps periodically to fight drift

Pipeline (condensed)

ASR with word-level timestamps

Segment cleanup + merge tiny fragments

MT twice → produce A/B subtitle tracks; constrain length to reduce overflow; have multiple language model to supervise/double check the translation quality

TTS → single voice (female/male) for the full track

Mixback → keep ambience, duck original, SRT (mono or dual-lang) + dubbed MP4

Why post this now It’s not “magic studio” quality, but it’s dependable for many real-world cases: course videos, onboarding, webinars. We found being explicit about limits (no diarization) actually speeds teams up.

What we’d love feedback on

The current User experience

Acceptable subtitle overflow rate on 30–60 min content

TTS pacing rules that feel most natural for multilingual reading speed

(If there’s interest, we’ll extract a minimal CLI with the exact steps and corner cases called out.)

Comments URL: https://news.ycombinator.com/item?id=44893553

Points: 1

# Comments: 0

Show HN: EchoKit –Fully Open Source AI Voice Agent Hardware and Server

3Sophons — Thu, 17 Jul 2025 17:45:04 +0000

It is powered by Rust most “open” AI devices are actually closed on the backend, limiting customization and privacy. EchoKit is fully open—from ESP32-S3 hardware to the Rust-based server. You can customize ASR, TTS (including voice cloning), LLM prompts, and even add your own MCP actions. We’d love your feedback, ideas, and contributions. Try it out, check the code, and let us know what you think!