@Akbar501

Building with AI

 Palo Alto, CA
sentrix.ai
Joined June 2012
  • Tweets2.9K
  • Following477
  • Followers201
26Media
Everyone asking what’s the use case when it’s obviously getting pricing when a website says schedule meeting to get pricing.
Tavus
@tavusOct 1
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…
18
Absolutely intense competition in AI personal agents: Claude Cowork: Jan 12, 2026 Vellum: Jan 13, 2026 Hermes: Feb 25, 2026 Instinct: Aug 2026 Grokbot: Aug 11, 2026 Meta Muse: Sep 8, 2026 OpenAI Dots: Sep 29, 2026 No doubt more to come before year end.
59
Just completed an interview in Claude. It's an interesting format. Much more natural than traditional survey questions.
27
Used @muse to turn some photos of an old bike into a marketplace post. It was definitely a lot easier and removed friction. As someone who struggles to sell old stuff, I can see where I’d use this.
27
Personal Agents will hit a wall with about half of consumers until 2FA is solved.
16
Just used Astra for UI design / frontend dev for the first time. Impressive.
35
What's the latest research on quality degradation as content usage increases? In the past, model quality dropped as content usage increased. Is that still true?
1
27
What's the best effort level for generating UI mockups with Astra?
1
30
To everyone using up their entire Claude subscription with a few Fable 5.1 prompts. What are the prompts that you’re using? And at what effort level?
38
Early results are promising with Fable 5.1. Code quality good. Pleasant to work with.
33
The whole AI can build anything argument relay highlights who does not understand the difference between software vs product vs a business.
70
The whole one shotting an application is impressive, but not worth being impressed by. Models surpassed this ability a while ago. And still a production application requires more logic and thought than can be fit into a single prompt.
1
1
87
Coding agents will overcomplicate things if you don't watch them closely.
1
1
76
The current AI models are absolutely excellent at one-shotting display only apps that have minimal logic. It's actually nuts how quickly you can create a dashboard that summarizes some information that you're studying.
55
If all you have is a hammer, everything starts looking like a nail.
39
I still find that the latest models fail to complete all requirements when the implementation is complex/nuanced. 9 times out of 10 there needs to be an iterative review cycle to get all requirements implemented.
41
When doing a code review with Claude/Codex, instead of asking for a traditional LoE, ask it what effort level is needed to fix each issue. That way you can use your tokens and time efficiently by using the right model for each issue.
1
2
107
Did OpenAI just do another reset. Awesome.
29
If everyone is a founder, who's left to be the user? If everyone is an investor, who's left to be the founder?
27
Hey @ClaudeDevs Thank you for this. Very generous. I will use it wisely.
20
Load more