
Building @cerebriumai - making it easy to deploy real-time AI at scale. Prev exit to Walmart. Kind of a big deal
New York
Joined October 2011
- Tweets306
- Following225
- Followers248
- Likes88
Introducing Griffin, the first model to pass the video Turing test.
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
It’s the first Human Interaction Model (HIM).
Readers added context they thought people might want to know
The 48% figure and "video Turing test" claim are from Tavus's own study of 54 one-minute calls, not independently verified or using a standard protocol. Griffin-Lite leads NVIDIA's VideoFDB benchmark on their public leaderboard.
cellcog.ai/blog/tavus-gri…
research.nvidia.com/labs/amri/proj…
tech-ish.com/2026/10/02/tav…
Top 3 worst experiences:
1. Flying
2. Being a passenger in a electric taxi - the constant acceleration and braking
3. The GCP UI
2
74
Michael Louis retweeted
Today we’re sharing how we built Thalamus: Cerebrium’s highly available distributed router for global AI workloads.
It helps power low-latency, resilient AI applications across fragmented compute for companies like
@tavus, @superwhisper, @useCamb_AI
Read more: cerebrium.ai/blog/thalamus-o…
This is awesome to see! I always wondered about the appetite for Governments in Africa for an @anduriltech type company.
These Gen Zers just raised $11.75M to put Africa’s defense back in the hands of Africans techcrunch.com/2026/01/12/th…
1
281
Most voice AI teams delay this decision far longer than they should - and eventually hit the same wall.
The LLM ends up being the single biggest driver of both latency and cost.
3
124
This typically only starts to make sense at ~15k+ calls per day.
But that’s also when customers start noticing.
Lower latency shows up immediately in:
• Conversation flow
• Turn-taking
• Perceived intelligence
1
1
71
One of the more interesting AI reports I’ve read recently is @openrouter's State of AI - because it’s based on real production usage at scale, not benchmarks or demos.
A few takeaways 🧵👇
You can read the full report here: openrouter.ai/state-of-ai
2
2
84
5. Open-source models now represent ~30% of total usage.
Driven by:
• Cost
• Flexibility
• Faster iteration cycles
I’d be surprised if this number doesn’t grow meaningfully in 2026.
6. Early model fit creates lock-in.
When users find a model that fits their workflow early, they stick with it.
1
40
The bigger signal....
The winners won’t just be the models with the best benchmarks - but the platforms that deliver:
• Low latency
• Reliability
• Cost efficiency
• Great developer experience
• Production-ready infra
Infrastructure is the moat.
And that’s exactly what we’re focused on solving at @cerebriumai .
39
Infra matters. xAI built Colossus, one of the world’s largest AI supercomputers, in record time.
Outside of Google, very few labs own their full stack - OpenAI and Anthropic still rely heavily on cloud partners.
Distribution + data are underrated advantages.
xAI already reaches ~600M active users via X, competitive with ChatGPT at ~900M, with a still-nascent GTM - plus proprietary real-time data no one else can access.
1
34



