ElevenLabs $500m Series D at $11B, Cerebras $1B Series H at $23B, Vibe Coding -> Agentic Engineering
Summary
The AI news brief for February 3-4, 2026, highlights significant developments across the AI landscape, including major funding rounds for Cerebras and Eleven@11, with Cerebras achieving a "DOUBLE decacorn" valuation of $23B. Google is integrating Gemini 3 into Chrome and reporting rapid adoption with 750M+ MAU for the Gemini app and a 78% unit-cost reduction for serving. Coding agents are converging in IDEs, with VS Code introducing "Agent Sessions" and GitHub Copilot supporting Claude and OpenAI Codex. New benchmarks like METR's GPT-5.2 (high reasoning effort) show a ~6.6-hour time horizon for software tasks, while Perplexity released its DRACO benchmark. Multimodal generation sees advancements with Kling 3.0's custom multishot control and improved detail, and Grok Imagine gaining momentum in video generation arenas. Research notes cover LLM reasoning, continual learning, and robotics with "World Action Models."
Key takeaway
For CTOs and VPs of Engineering evaluating AI integration strategies, the rapid productization of models like Gemini 3 and the convergence of coding agents in IDEs signal a maturing ecosystem. You should prioritize platforms offering robust agentic workflows and consider the cost efficiencies of specialized hardware like NVIDIA GB200 for inference, while also scrutinizing benchmark methodologies for "economically useful work" to ensure real-world applicability.
Key insights
AI development is accelerating, marked by massive investments, widespread model integration, and evolving evaluation methods.
Principles
- Agentic engineering is emerging as a new meta-paradigm.
- Observability shifts from stack traces to trace-driven evaluation for agents.
- Continual learning is crucial for building robust agent memory.
Method
Google is benchmarking "soft skills" by having models compete in games like Poker and Werewolf via the Kaggle Game Arena, testing planning and decision-making under uncertainty.
In practice
- Integrate Claude and Codex agents into VS Code via Agent Sessions.
- Utilize vLLM on NVIDIA GB200 for 3-5x throughput gains.
- Explore Kling 3.0 for multi-shot video generation with native audio.
Topics
- AI Industry Funding
- Large Language Models
- AI Agents
- Multimodal AI
- AI Benchmarking
Code references
Best for: Investor, CTO, VP of Engineering/Data, AI Engineer, Machine Learning Engineer, AI Product Manager
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by AINews.