frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: ThunderPhone v2 – a new architecture for voice AI

https://thunderphone.com
4•kolchinski•1h ago
Hi folks, Alex here from ThunderPhone. Today we're launching v2.

Voice AI has been a hot topic for the past year or two, both for automating phone calls and for adding voice features to apps.

Most of the voice agents deployed have been using a "3 step pipeline": a transcription model to turn user speech into text, an LLM to generate a response in text, and a TTS model to turn the response text into audio.

But the performance of these 3-step voice agents has often been so-so, largely for 3 reasons:

1) Voice agents have to be fast, usually requiring non-thinking LLMs to handle the conversation. Non-thinking LLMs make mistakes, leading to dumb behavior that derails conversations.

2) Today's voice AI stacks typically rely on a single transcription model to turn what the user says into text. This step loses a ton of information from the audio, and if the transcription model makes a mistake, the LLM often has no way to recover. This also leads to a lot of dumb behavior that breaks calls.

3) Natural conversation handling is a hard problem - filtering out noise like background voices, knowing when to allow the AI to be interrupted by a "uh, wait" but not by an "uh-huh", etc. - this also leads to awkward conversations.

The "bitter lesson-pilled" solution to all of these is most likely going to be a "full duplex" model that receives and emits audio at all times, allowing for fluid back-and-forth, while also calling a smarter model behind the scenes. OpenAI appears to have been the first to make real progress towards this architecture with their latest GPT-Live release, but that tech is not yet ready to plug into phone calls.

In the meantime, we've stitched together a stack that improves the performance of phone calls far beyond what's possible with a 3-step pipeline.

The ThunderPhone stack combines a few insights:

1) Grabbing signal from audio in more than one way, including running multiple transcription models at once, and piping audio directly into LLMs. This hugely reduces mistake rates, especially on challenging problems like data entry, multilingual calls, etc.

2) Combining thinking and non-thinking LLMs: in a conversation, it's natural to respond to some things more quickly than others, and sometimes even say things like "oh, let me think about that" - to indicate that it'll take longer to get back to someone with a final answer. ThunderPhone does the same thing.

3) This is less unique to us but we've stitched together a very big swarm of small (and in a few cases large) models to help make conversation handling more natural, even in hard environments like loud places, speakerphone, etc.

We've made the ThunderPhone stack available at 3 price points - 2c/min, 5c/min, and 9c/min, each with their own level of capability.

The 2c/min model ("Spark") is the cheapest on the market to our knowledge, and is smart enough to handle simple transactional calls.

Bolt at 5c/min is a middle ground, and the fastest model we offer.

Storm at 9c/min (+3c/min for extra intelligence) is our flagship model, able to handle even quite complex calls with very few mistakes. With extra intelligence turned on, it sets the record on the Big Bench Audio benchmark at 99.4% accuracy.

ThunderPhone is mostly aimed at B2B applications, but is also useful if you want to do something like setting up a smart voicemail for yourself, or calling around restaurants to make a reservation, calling around pharmacies to find a prescription, etc. for personal use.

Let me know if you have questions - happy to answer anything about ThunderPhone except a few "trade secret" specifics, and about the voice AI space more broadly, including where we think the tech is going.

And comment here or email me at alex@thunderphone.com if you want to get set up with some credits to try it out.

Alex

Dyna-2: A 1M-Hour Scaling Law for World-Action Models

https://www.dyna.co/dyna-2
1•gmays•1m ago•0 comments

US Says CN Hackers Broke into Justice Department, NASA, Federal Reserve, Senate

https://www.reuters.com/world/china/china-sponsored-hacking-platforms-seized-by-us-justice-depart...
1•Teever•1m ago•0 comments

Show HN: ASOGrade – App Store keyword research using Apple Search Ads demand

https://asograde.com
1•ashwinn11•1m ago•0 comments

Euphemisation of Taboo Areas [pdf]

https://ajmp.uwr.edu.pl/wp-content/uploads/sites/39/2023/12/10_Kleparski.pdf
1•Eridanus2•3m ago•0 comments

Vast Space – the artificial gravity bet that could unlock deep space

https://beyondhorizonforesight.substack.com/p/vast-space-the-artificial-gravity
1•beyondhorizonfs•4m ago•0 comments

Teaching AI to Read a Map

https://research.google/blog/teaching-ai-to-read-a-map/
1•mattrighetti•5m ago•0 comments

Ultraspherical

https://www.johndcook.com/blog/2026/08/26/ultraspherical/
1•ibobev•7m ago•0 comments

Numerical (In)Stability of Recurrence Relations

https://www.johndcook.com/blog/2026/08/24/numerical-instability-recurrece/
1•ibobev•7m ago•0 comments

Three-Term Recurrences

https://www.johndcook.com/blog/2026/08/24/three-term-recurrences/
2•ibobev•7m ago•0 comments

Mark Zuckerberg Wants to Make Sure His Competitors Share His Pain

https://www.nytimes.com/2026/08/27/technology/meta-settlement-mark-zuckerberg-youtube-tiktok.html
5•aaraujo002•9m ago•1 comments

Why Did Stripe Acquire an AI Model Routing Company?

https://thefinancialengineer.substack.com/p/why-did-stripe-acquire-an-ai-model
3•gemanor•11m ago•0 comments

Lockstep – Can a language model run a logic circuit in its head?

https://lockstep.greg.technology/
2•gregsadetsky•12m ago•0 comments

Formalization of the Solution to the Hopf Problem

https://github.com/plby/HopfProblem
2•robinhouston•12m ago•0 comments

You are not a model. Don't price per token

https://twitter.com/sarahdingwang/status/2092976778164011481
2•tosh•13m ago•0 comments

A phone call that saves lives

https://www.gatesnotes.com/heroes/heroes-home-topic/reader/heroes-in-the-field-m-mama
2•mooreds•13m ago•0 comments

Agents still can't automate Excel

https://www.orcaset.com/blog/agents-still-can-t-automate-excel
3•jrdnh•14m ago•0 comments

Show HN: IdeaKache – Ideas Worth Holding Onto

https://ideakache.com
3•jcowcher•15m ago•1 comments

Loop Engineering

https://addyo.substack.com/p/loop-engineering
2•fagnerbrack•16m ago•0 comments

Why HPSC Is a Big Deal for Space Exploration

https://www.windriver.com/blog/Why-HPSC-Is-a-Big-Deal-for-Space-Exploration
2•mooreds•16m ago•0 comments

Markdown Database Pattern

https://wayofmarkdown.com/markdown-database
3•rufuspollock•16m ago•0 comments

Codefrens – Two strangers. One problem

https://codefrens.com/
2•dyftvugj•17m ago•0 comments

Want to listen better? Turn down your thoughts and tune in to others

https://text.npr.org/974786825
4•mooreds•17m ago•1 comments

Apple upgrades M4 Mac mini orders to M6, M5 Pro at no additional cost

https://www.macrumors.com/2026/08/26/apple-free-mac-mini-upgrade-email/
3•alwillis•19m ago•1 comments

Small Models Have Arrived

https://calv.info/small-models-have-arrived
2•tosh•20m ago•0 comments

Cocomelon's Studio Tells Its Artists to Start Experimenting with AI

https://www.404media.co/cocomelons-studio-tells-its-artists-to-start-experimenting-with-ai/
2•cdrnsf•20m ago•0 comments

Europe's Heatwaves Are Putting Nuclear Power Under Pressure

https://oilprice.com/Energy/Energy-General/Europes-Heatwaves-Are-Putting-Nuclear-Power-Under-Pres...
4•toomuchtodo•21m ago•1 comments

Thailand Accelerates Clean Energy Push to Cut LNG Dependence

https://oilprice.com/Latest-Energy-News/World-News/Thailand-Accelerates-Clean-Energy-Push-to-Cut-...
3•thelastgallon•21m ago•0 comments

The broadcast squeezeback, rebuilt with CSS Grid and WebVTT

https://www.mux.com/blog/the-broadcast-squeezeback-rebuilt-with-css-grid-and-webvtt
4•mmcclure•21m ago•0 comments

Suica, Japan's First IC Transit Card

https://www.tokyodev.com/articles/the-story-of-suica
9•zdw•22m ago•4 comments

Japanese polka dot artist Yayoi Kusama dies aged 97

https://www.bbc.com/news/articles/c3v4k0re3vwo
45•herbertl•23m ago•4 comments