tmshapland

Thank you for writing this thoughtful post, Armin. I find it deeply comforting that the developer of Pi, an agent harness, does not remove himself from the loop, like me. Maybe if I started thinking of codebases as biological organisms I could get comfortable with getting the human out of the loop.

Thank you. 100%.

How do we make this site mainstream? The public would really start to push back if they could so viscerally experience that they are being surveilled multiple times per day.

Congrats the launch, guys!

Here's the link to the LiveKit LemonSlice plugin. It's very easy to get started. https://docs.livekit.io/agents/models/avatar/plugins/lemonsl...

lol. would you share the prompt for how you translate them? it really feels like a snarky HN community member rewrote each one.

Beautiful. Reminds me of starling murmurations. https://www.youtube.com/watch?v=V4f_1_r80RY

Fascinating! How did you decouple the speaker-specific vocal characteristics (timbre, pitch range) from the accent-defining phonetic and prosodic features in the latent space?

These avatars from Hedra inspire the artist in me, the self that craves something weird and intuitive and unfathomable, something that appeals to the emotional self. The internet's most popular sites are polluted with rational self-promotion, where everything has its aim (likes, impressions, sales). Must everything be up and to the right?

Such a fascinating read. I didn't realize how much massaging needed to be done to get the models to perform well. I just sort of assumed they worked out of the box.

Sample code for trying our gpt-oss as the LLM in the voice agent stack.

here's how it performs as the llm in a voice agent stack. https://github.com/tmshapland/talk_to_gpt_oss

Congrats on joining Mintlify, friend. Trieve is dead, Long live Trieve!

Man, I've needed this! Brilliant!

I am a former YC founder who works at LiveKit now. I love it here! So much opportunity! Such a talented, hard working, and collaborative team!

yes, so true! Even if they've been spending around that much every year, it's still an amazingly good ROI for Intuit to pay off lawmakers.

It's amazing how little it costs to pay off lawmakers. I've seen this before in the first Trump admin where some private weather services paid a fairly small amount (tens of thousands) to get some NWS services shut down. Considering how much Intuit has to gain, they really didn't have to pay much. Great ROI.

"A glance at Intuit’s 2025 first-quarter lobbying disclosures gets at this continued, quarter-century saga. The company shelled out $240,000 to lobby members of Congress on tax-related issues."

This is cool! Thank you for responding with this! Really neat to see the K2-18 spectra too.

I've been waiting to see atmospheric composition results from JWT. It seems like JWT was first put to work on other tasks. Anyone know of other studies from JWT like this?

Exactly.

Sad.

Awesome! We've been waiting for Krisp's voice isolation on the server side in the Voice AI community! This is an important advancement for the community! Congrats, Krisp team!

Thank you for posting this.

For Radiohead’s 1997 ‘Fitter Happier’, Thom Yorke used the MacInTalk speech engine’s ‘Fred’ voice to produce the vocals. We updated the vocals with ElevenLabs modern speech synthesis.

Here’s a link to the code.

https://github.com/Canonical-AI-Inc/canonical-voiceover

"So we’re going to make a damn iOS app." Love it.

Here is a public link to the paper.

https://drive.google.com/file/d/1QOTn2gme6AlfCnfmFa-gXSrhuFO...

I've always wanted to be able to update what I believe and feel about something like I can update code. This method has worked for me, although it's more like training a neural network model than editing a file.

The novel part of the method is in how it separates noise from the recurring shape in the time series. In the paper, they are separating random noise from recurring ramp shapes. But the same procedure could be implemented for other recurring asymmetric shapes in time series.

The method is really good at getting the characteristics of the recurring shapes. In the case of ramps, it's good at getting the average height of the ramps and the average duration of the ramps.

This is so cool! Thanks for sharing! Would you mind pointing me to where I could listen to some of the sounds produced by analog electronic synthesizers running ramp functions? I'd love to hear it. I looked through the repo and didn't find an easy way to listen.

Often it seems that old papers are the most insightful. Back then, the research cycle was shorter and scientists had more institutional support. They had more time to think.

This paper has always amazed me. It's an incredibly creative means to getting the characteristics of a recurring shape from a noisy time series signal. It's sort of like what wavelet analysis does. It's remarkably effective.

The author skips a lot of steps in his work in the paper. Here is a link to a stepwise derivation of the method.

https://drive.google.com/file/d/1ytbqe-zL9j7hddm4TTU3egXdnm7...