All Model Labs Are Now Agent Labs (www.latent.space)

AI21 has officially rebranded its Model Labs to Agent Labs, shifting its focus from traditional model development to harnessing AI agents. This transition underscores a significant evolution in the AI/ML landscape, validating the stance that effective AI solutions now require a synergy between models, harnesses, workflows, and user interfaces. As organizations like DeepSeek pivot to building dedicated teams for agent development, the implications are profound: the interplay between model training and post-training with harnesses may create tighter controls over model usage, potentially leading to more closed ecosystems that could limit broader API competition. As the industry evolves, the emergence of agent products emphasizes that mere model performance is no longer sufficient. Recent discussions highlight that the future of AI solutions lies in integrated systems where models work symbiotically with harnesses and workflows. OpenAI's latest updates to their Codex demonstrate this trend, as new features enhance productivity while stirring conversations around the balance of cost and performance. The reinstated emphasis on comprehensive product ecosystems marks a pivotal shift, suggesting that the AI/ML community must adapt to a landscape where collaborative, multi-faceted solutions become the key to success.