Back to collectionAgent Work Expands Across Research, Tools and Incentives
Virtuals and Kaito add distinct activity to NEAR's confidential-compute developments.
Virtuals published a task-market report, introduced gold and equity research agents and discussed robotics around Unitree. NEAR's parallel developments included staking-linked confidential inference across more than 40 models and IronClawAI 1.2 support for Slack, multi-user agents and custom MCP connections. Task reporting, research tools and computing access represent different stages of agent work.
Kaito Studio connected with Megapot around creator rewards and recruited an initial Katalyst group, while Pulse raised privacy and social-interoperability questions. Chainlink's CCIP support for Wyoming's FRNT stablecoin placed another named value-transfer service alongside these developments. None of these connections alone establishes an agent's authority to spend.
The analytical link is the handoff from information to consequence. A research agent can inform a decision without controlling payment; a reward program can attract contributors without proving demand for their output. Combining those functions makes provenance and permission boundaries more important, rather than turning activity across several services into a single measure of autonomous adoption.