How Will CoreWeave Forge Help Enterprises Continuously Refine AI Agents in Production?

CoreWeave Inc. has launched Forge, a full-stack platform designed to accelerate the continuous improvement cycle for agentic AI. According to SiliconANGLE, the offering aims to help enterprises move from initial agent deployment to optimized performance more quickly by unifying disparate tools into a single workflow.

Why is the AI improvement loop considered continuous?

Enterprises deploying AI agents are finding that development does not end at launch. Agents must be run, observed, evaluated, and improved on an ongoing basis. CoreWeave vice president of product marketing Susanne Seitinger told theCUBE Research that Forge provides a shared workflow for business units and machine learning teams. This structure allows different groups to communicate more effectively and treat each production run as an opportunity to refine the model.

What are the specific stages of the Forge workflow?

The platform structures the improvement process into five distinct stages: run, observe, curate, improve, and evaluate, before repeating the cycle. Each pass is intended to build upon the previous one. Seitinger stated that the speed of learning is the primary metric of success. Faster learning leads to faster shipping and quicker value realization, which supports the broader cloud infrastructure CoreWeave is building.

Which tools within Forge address specific development challenges?

Several new capabilities target difficult parts of the cycle. Agent Lens allows teams to inspect agent behavior, while Registry models checkpoints and configurations. RL Rollouts and model distillation handle the improvement phase. Additionally, CoreWeave’s AI Research and Iteration Agent, or ARIA, is now generally available to help users spot patterns. Seitinger noted that distilling knowledge from large models into specific use cases is critical for organizations.

How does this launch align with current hardware trends?

The timing reflects a rapid shift in workload mixes. Cognition AI Inc. is already running production workloads on CoreWeave’s first Nvidia Corp. Vera Rubin NVL72 racks. This activity underscores the reality that post-training and continuous improvement are now standard operational requirements. The platform is designed to support these ongoing processes rather than treating model development as a one-time event.

Background: Agentic AI

Agentic AI uses a large language model as a reasoning engine to plan, call tools, and observe results in a loop. It does not just generate text; it executes functions. You control it by defining available tools, setting strict output schemas, and monitoring the feedback cycle for drift.

Read the full guide: How Agentic AI Works: The Loop, The Limits, And The Leak

What to do and how to stay safe: CoreWeave

• Review your current agent evaluation metrics to ensure they capture continuous performance drift.
• Identify which teams currently handle observation versus improvement and assess communication gaps.
• Monitor vendor announcements for new tools that support model distillation and configuration registry.
• Plan for post-training workflows as a standard part of your deployment lifecycle.

Step-by-step guide: How AI Risk Management Works: The Hidden Failure Points

General security guidance from the Sandbox Dispatch newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What is CoreWeave Forge?

Forge is a full-stack platform from CoreWeave designed to give teams a shared workflow for running, observing, and improving agentic AI models continuously.

Is the ARIA agent generally available?

Yes, CoreWeave’s AI Research and Iteration Agent, known as ARIA, is now generally available to help users spot patterns in their data.

Which company is using CoreWeave’s new hardware?

Cognition AI Inc. is already running production workloads on CoreWeave’s first Nvidia Vera Rubin NVL72 racks.