Alex Kolchinski

@phodaie @juberti +1, including intelligently tolerating background noise, quiet/loud speech, overlapping conversations etc

@juberti Semantic VAD and other future improvements in VAD

Let me know what topics you're most interested to hear about tonight!

@samuel_spitz Yeah I've had SF people tell me voice AI was over lol...

It's crazy how much the startup talk differs between LA & SF I told someone in LA I was building AI voice. They said "AI phone calls are kind of slept on" I told someone in SF I was building AI voice. They said, wow you picked literally the most competitive space possible.

@koomen The other way in which I think we're still in "horseless carriage" land is having to read/type and interact with static software vs. talking to computers and having content be dynamic, made a demo around a new vision of this at talktosales.com

I wrote an essay about Gmail’s useless email-writing AI assistant: “AI Horseless Carriages”. link and TLDR in thread

@knowclarified we use voice-to-voice at ThunderPhone.com, are you trying to do phone, web, or both?

Anyone have experience working with OpenAI’s realtime, or any voice-to-voice ai applications? Pls comment. I have questions

@kwindla Not yet but soon!

@kolchinski Can you talk about what speech-to-speech model you are using, or if you trained it yourself, the architecture? So much interesting stuff happening in audio/speech models, these days.

Voice agent cost calculator (and a couple of common calculation mistakes) Here's a spreadsheet to calculate the per-minute cost of a voice agent. I cleaned this up to share because I've had the same surprising conversation a few times lately ... 🧵

@kwindla ThunderPhone.com is only 2 cents a minute! No self-hosting required.

Several people have asked why voice agent platform costs seem to be anchoring around $0.10/minute, if the per-minute cost of of using APIs directly and hosting code yourself is $0.02-$0.04/minute ... Part of the answer is that "hosting code yourself" has more moving parts than just paying for vCPU runtime cost. (The cost calculator linked in the thread below only includes the vCPU cost, benchmarked as AWS on-demand instance cost.) You'll also need at least: - service discovery - load balancing - logging - monitoring - bandwidth At scale, you'll need: - more of all of the above :-) - compliance - multiple regions - analytics - customer support A voice agent platform has to pay the hard costs for all of these, plus the salaries of the people doing devops. There are some ways to push costs down below $0.02/minute, though today these require making quality compromises, operating at very large scale with an experienced devops team, or having a specific use case that's different from what most people are doing in the voice AI space. In general, it's very hard to run an LLM cheaper than Google will serve you Gemini Flash tokens! It's also hard to run a speech model that's cheaper than Cartesia or Deepgram, once you factor in provisioned GPU capacity. This whole space is evolving rapidly, though. There's some cost pressure on providers, as many commercial and open source models continue to get better and better.

What if websites were voice-controlled?🤔 That's what we're launching today with TalkToSales, which adds AI assistants to websites. 🤖 They answer questions, give live tours, and even book immediate sales — boosting conversion rates.📈 We're onboarding our first customers now.

@geoffintech Best preparation of all, just start the company!

There is no better preparation to become a founder than to become a PM at a fast growing startup. Excited to rank #4 here in just year 5 and to continue building an incredible team that will accomplish great things - at Ramp and beyond.

@CCgong Sounds like they're not offering enough equity 😉

Startups who raise at too high of a valuation risk not being able to hire the best talent (or raise their next round). Just chatted with cracked engineers who rejected offers from the “hottest” AI startups because their valuation was too high - “what’s my upside here?”

@moose_antler @jsnnsa @Itsjoeco +1, not a parent and this is a strange opinion

@jsnnsa @Itsjoeco 31/32 now. Imagine never meeting your kids because some VCs were too optimistic. Couldn't be me.

having kids in the next 5 years might be a tragic mistake every smart bio founder/scientist i’ve talked to seems to think embryo editing for things like short sleeper, reduced cancer risk, etc is possible on a near term horizon imagine having two kids a couple years apart, one is a super human and the other isn’t “sorry jim, little timmy won’t get cancer, only needs 4 hours of sleep and you’re normal you were just born in the wrong order 🤷🏻 “

@paulg The Infatuation is great in SF, hopefully in other cities too

Is there a reliable source of restaurant ratings, like Zagat's used to be?

@HarryStebbings @HarryStebbings what patterns if any are you seeing re industries/type of company (size, niche etc.) and use cases with fastest big adoption? Especially curious if it's primarily software companies adopting heavily first vs. ones in other industries.

Everyone who says AI is in the experimental budget phase of enterprise spending does not see it. Enterprises are experimenting with AI but where it works and provides sufficient value, they are spending BIG. Seen 7 companies scale to $1M contracts in the last 30 days.

@NNNIncome Just read Men to Match my Mountains (a history of the American West) and SF was briefly majority Mormon! And there were a lot of Mormons in SoCal too. But then most of them moved to Utah.

Every time I land in San Jose my first thought is, “Why did we stop the wagon trains in the desert Brother Brigham?” I mean it’s just stunningly beautiful and then side bonus it’s Silicon Valley so why didn’t we settle here instead??

@andrewoetting @jaltma Which analogies from the cloud era come to mind? I'm thinking about this a lot too and could use the history lesson.

@jaltma Agreed, two sides to the coin Lots of wedges start up take for granted will close And lot of new “fools gold” wedges where the gaps will be filled by incumbents (increasing stickiness) vs opportunities for start ups Many analogs to this from cloud compute wave

Jotted some thoughts about how AI opens up new killer app possibilities which can be wedges to systems of record

It's interesting how similar San Francisco's problems were in the 19th century as now — residents so busy making money that they ignore politics, leaving politicians free to feed at the public trough and ignore public safety.

I had a chance to productionize ReadToMe while shutting down my previous company, and I'm launching the publicly-available version today. Give it a try if it sounds useful to you, and please let me know if you have any feedback.

I wrote it initially as a Christmas present for my fiancée, who has a reading disability. She loves audiobooks but many books don't exist in audiobook form and existing apps are surprisingly poor at the paper-to-audio use case.

Launching ReadToMe today — turn paper books into high-quality audio with your phone camera. readtome-app.com

@willdepue Braudel’s Civilization and Capitalism part 1 is an incredibly wide ranging and insightful view of immediate pre modern European society

who do you know/what book have you read that has the most comprehensive understanding of how to model the world?