Show: Google Cloud Next · Publisher: Google Cloud · Host: Google Cloud editorial
Episode URL: https://www.youtube.com/watch?v=er08iXtLIxI
Publish date: 2026-04-14
Duration: 175.0s
Default source credibility: MEDIUM — Vendor conference — Google keynotes and product launches are marketing (LOW for claims about own products). Named F500 customer talks with production metrics stay HIGH. Filter aggressively for customer sessions with named speakers and quantified outcomes; skip product demos and partner pitches.
- Replit’s Rapid Agent 3 can run autonomously for 200 minutes, enhancing productivity for knowledge workers.
- Google Cloud provides reliable infrastructure for Replit, enabling 24/7 availability and scalable deployment.
- Replit uses a mix of advanced and cost-effective AI models to balance intelligence and affordability.
Extracted quotes
| # | Credibility | Speaker | Org | Timestamp | Topic | Quote |
|---|---|---|---|---|---|---|
| 1 | MEDIUM | Mikall Katasta (President Head of AI) | Replit | 0:39 | 01-ai-native-landscape | It’s capable of running for 200 minutes in a row. It’s capable of testing the applications so that a user is just left with expressing natural language what they want to build. |
| 2 | MEDIUM | Mikall Katasta (President Head of AI) | Replit | 1:29 | 02-corporate-tools | We also use a lot of the AI models offered by Vert.Ex. We also use Gemini models extensively. Not only the frontier ones that of course are very important to bring in the right level of intelligence and capability to the agent, but we also use cheaper models that are faster and there we flooded some of the side task that the agent has to accomplish that allows us to keep the agent very affordable and also at the same time make it move faster. |
Per-quote detail
1. Mikall Katasta — Replit (0:39)
It’s capable of running for 200 minutes in a row. It’s capable of testing the applications so that a user is just left with expressing natural language what they want to build.
- Stat: 200 minutes of continuous operation
- Credibility: MEDIUM — Specific metric with duration, but vendor context.
- Topic tag:
01-ai-native-landscape
2. Mikall Katasta — Replit (1:29)
We also use a lot of the AI models offered by Vert.Ex. We also use Gemini models extensively. Not only the frontier ones that of course are very important to bring in the right level of intelligence and capability to the agent, but we also use cheaper models that are faster and there we flooded some of the side task that the agent has to accomplish that allows us to keep the agent very affordable and also at the same time make it move faster.
- Credibility: MEDIUM — Specific use of AI models, but vendor context.
- Topic tag:
02-corporate-tools
Extracted 2026-04-14T12:54:12 via scripts/podcast_mine.py (MLX mlx-community/Qwen2.5-32B-Instruct-4bit).