San Francisco, CA
Joined August 2015
- Tweets2.9K
- Following1.5K
- Followers2.8K
- Likes57.8K
kabir retweeted
This festive season, we're excited to share that Nugget by @zomato and Cartesia have been powering millions of customer interactions through multilingual voice agents for enterprises in retail, BFSI, healthcare and more.
Nugget is a conversational AI ecosystem that has powered 2.2B+ conversations for brands including Zomato, Blinkit and Hyperpure, and handles 50M+ minutes of voice interactions every month.
Cartesia's flagship text-to-speech model, Sonic, powers Nugget's agent voices, bringing natural, expressive, and low-latency speech to conversations at scale across 11 Indic languages.
We’re proud to be helping customers throughout India connect with brands in their own languages, this festive season and beyond.
kabir retweeted
Gina Hinojosa just became the Governor of Texas. Dems may flip the legislature.
Pour one out for all those corporations running to Texas so they could escape the liberals in Delaware and New York…
For the normies (like me): Our new architecture is able to understand the basic structure of raw data *on the fly*, without relying on a fixed tokenizer!
For example, it learns to tokenize code in unseen languages after seeing it a few times.
It requires 10X less data and compute to converge to the same level of performance as a conventional model.
Especially exciting for domains we *don’t understand well*, where current models impose priors that fight the bitter lesson.
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
Cartesia’s website says, “Models and agents built on frontier research.”
Our H1’s don’t lie!
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
3
28
1,475
kabir retweeted
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
kabir retweeted
Cartesia, the superintelligence company?
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
kabir retweeted
this chart is kind of crazy no
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
kabir retweeted
Cartesia Managed Agents is #1 on the @ArtificialAnlys Speech Agent Arena, outperforming 20+ speech-to-speech systems on tasks like appointment booking and ordering.
We achieve frontier quality and naturalness by co-designing our speech models and harness in one vertically integrated stack.
Cartesia Managed Agents scores an Elo of 1115 on the Artificial Analysis Speech Agent Arena, 98 points above Cartesia's earlier voice agent product, Cartesia Line, with Task Success Rate up from 77.5% to 91.7%
Managed Agents is @CartesiaAI's fully managed platform for building and deploying voice agents. It is a cascaded system: Cartesia's own Ink Speech to Text and Sonic Text to Speech models run around an LLM of the developer's choice. We evaluated Cartesia Managed Agents with GPT-5.5 as the LLM and the Skylar and Daniel voices.
Key takeaways:
➤ Speech Agent Arena: Cartesia Managed Agents scores an Elo of 1115 (95% CI 1086-1144, 685 samples), 98 points above Cartesia Line at 1017 (95% CI 985-1049)
➤ Agentic tasks: On tasks that need tool calls, Cartesia Managed Agents scores an Elo of 1112
➤ Task Success Rate: 91.7% of eligible tool-calling conversations end with the correct final task-completing tool call
Congrats to the @CartesiaAI team!
See below for more details ⬇️
Given that all the AI companies owe each other money, I suggest they make a Venmo group and turn on the "simplify transactions" feature
4
58
2,154
kabir retweeted
Conspiracy theorists: the NSA has manipulated the standard setting process so HTTPS appears to be impenetrable, but they have a backdoor because of advanced mathematics beyond your imagination
The real world, so much dumber than they could ever imagine: the secret key is 4
kabir retweeted
if you're building harnesses for real work, this is a great read on how Ramp built the shared platform behind its agentic products like inspect, glass, etc.
We at Ramp tried a bunch of agent harnesses and ended up building our own. Cloud-native, model-agnostic, tailored to our infrastructure and product.
And we published how we built it: builders.ramp.com/post/build…
