Panos Stravopodis

The hardest problems in Voice AI can't be solved by Claude. (We checked.) The @Elyos_AI team will be in San Francisco next Monday hosting a dinner for Founders and Founding Engineers deep in this space. We've got an amazing group coming to talk about what's actually hard in voice right now. Come join us 👇 luma.com/zr3xlueb

@garrytan Love it! What about a /fundraise skill @garrytan? Might need it soon.

I just launched /office-hours skill with gstack. Working on a new idea? GStack will help you think about it the way we do at YC. (It's only a 10% strength version of what a real YC partner can do for you, but I assure you that is quite powerful as it is.)

It's Y Combinator's 21st birthday today.

If you geek out on real-time audio pipelines, follow along. Or even better: come build this with us 👇elyos.ai/careers?ashby_…

This screenshot is a reminder of how far we've come. Over the next few weeks I'll be sharing the models we tested, the latency tradeoffs, and what actually moved the needle in production. The stuff that doesn't make it into the docs.

"It doesn't meet our bar." That's what we said listening back to our early calls. The agent stepping on the user, missing cues, killing the flow. We were determined not to ship that. 🧵

@sukh_saroy Would be interesting to see how maintainable would that be.

🚨 ByteDance just open sourced a "brain" for AI agents. It's called OpenViking. A database that gives any AI agent real memory, real skills, and real knowledge. Right now, every AI agent forgets everything after each conversation. OpenViking fixes that. The idea is dead simple: store context like files on a computer. → Memories go in viking://user/memories/ → Skills go in viking://agent/skills/ → Resources go in viking://resources/ It also saves you money. Every piece of context has three levels: → L0: A one-liner (~100 tokens) → L1: The important stuff (~2K tokens) → L2: The full thing (only loads when truly needed) Your agent skims first. Digs deeper only when it has to. Here's the wildest part: After every conversation, it automatically learns from what just happened. No retraining. No manual updates. Your agent just gets better on its own. Built by the same ByteDance team running vector search behind TikTok since 2019. pip install openviking 1K+ GitHub stars. 100% Open Source.

The @Elyos_AI team will be in San Francisco next month and we are hosting a dinner for Founders and Founding Engineers building voice agents. Just good food, sharp minds and honest conversation about what’s actually hard right now in voice. luma.com/zr3xlueb

Don't miss out on this. This is a pivotal moment for founders to build something great. Deadline is February 9th. Apply now @ycombinator

WER by Turn Length

WER by Turn Type (e.g postcode, address, name etc)

At @Elyos_AI We benchmarked 13 STT providers on 100 real customer calls from the trades businesses. Not synthetic lab data. Real calls with: - Background noise & multiple speakers - UK postcodes & addresses - Regional accents (England, Scotland, Ireland) - Short confirmations to long explanations Top performers: 🥇 @DeepgramAI Flux - 15.9% WER 🥈 @soniox_ai - 16.9% WER 🥉 @Speechmatics - 17.7% WER @OpenAI Whisper? 39.8% WER - wouldn't recommend for production voice AI. What's your experience with STT models? Are we there yet?

@hasantoxr @hasantoxr have you seen any data from a production deployment? Anyone using it at scale?

🚨BREAKING: Microsoft just solved the "Agent Loop" problem. Agent Lightning is an open-source framework that lets agents learn from their own mistakes using Reinforcement Learning. Your agent fails a task → Agent Lightning analyzes why → Updates the prompt automatically → Next run succeeds. 100% Opensource.

In case you missed it, 2026 started with a bang for us

Crazy that you can actually run this locally. Will be sharing results soon.

We sat down with @pstrav, co-founder & CTO of @elyosenergy, to chat about building a climate tech startup why the power grid is broken, energy flexibility for commercial & industrial buildings, launching early, building things that don't scale, and more startuppirate.com/p/building-a-c…

Congrats @pipclairebrown, @adrianjohnst, @pstrav, and team @elyosenergy on the $3M seed! Elyos (YC S23) is on a mission to build a resilient and robust energy system by launching energy flexibility and demand response for mass markets. elyosenergy.com/blog/elyos-ann…

@KLM I am trying to get a refund for my booking as the second leg of my flight was cancelled and you rebooked me one with an extra leg adding 2 hours. I contacted your customer service and after holding for 45 mins they where unable to help. Can you please advise w/ next steps?