AI Agents & Web3 Agent Ecosystem

According to a recruitment listing from CMC Markets Polska, the company is hiring a Warsaw-based AI Growth & Performance Marketing Manager to build artificial intelligence agents capable of planning, launching, monitoring and optimizing digital advertising campaigns. The AI systems will automate

According to Chainwire, Ritual, an autonomous intelligence lab founded by former Polychain Capital partners Niraj Pant and Akilesh Potti, launched Ritual Chain on October 26, an EVM-compatible Layer 1 blockchain designed for autonomous AI agents. The network enables AI agents to operate as

According to Polygon, its Open Money Stack has enabled AI Agent spending controls, allowing teams to approve specific tasks, set budgets and expiration dates with wallet-level on-chain enforcement. Requests exceeding the budget, falling outside approved scope, or past expiration will be rejected

A panel discussion held at HSC Asset Management in Singapore on October 8, 2026, examined how stablecoins, blockchain infrastructure, and artificial intelligence could reshape global payment systems by 2030. Moderated by Declan Hannon of Aurora Labs, the event titled "The Payment Rails of 2030"

Oracle announced today the launch of Oracle AI Success Navigator, an artificial intelligence platform designed to accelerate the implementation and modernization of Oracle Fusion Cloud applications. The platform integrates AI agents, Oracle implementation expertise, and guided training to help

According to Robinhood's newsroom, Robinhood Ventures Fund I invested approximately $10 million in Cognition, the developer of autonomous software engineering tool Devin. The investment gives RVI shareholders indirect exposure to the privately held company through a publicly traded closed-end

According to Bloomberg's Odd Lots program, which aired on October 26, research from think tank Groundwork Collaborative suggests that some consumer-facing AI agents recommend more expensive products after learning about users' financial situations. Lindsay Owens, a researcher at the organization, st

CoinW launched 'AI Think Tank', a multi-agent AI analysis tool, on October 26. The platform-developed tool features 13 AI agents that simulate institutional research collaboration to analyze selected assets through data collection, investment debate, strategy formulation, risk review, and final

According to Nikkei News, Meta's AI agent Muse reached 10 million downloads within six weeks of its North America launch on September 8, surpassing growth rates of competitors including ChatGPT. The product entered Japan on October 23. Meta Chief AI Officer Alexandr Wang confirmed Muse will remain a

According to ChainCatcher, Google DeepMind researchers Zhengxuan Wu, Yuxuan Li, Oyvind Tafjord, and Been Kim released a preprint paper introducing XYEval, an evaluation framework designed to assess how AI agents respond to misleading user advice. The framework extends the concept of "XY problem" to

According to BlockBeats, CZ urged the Japanese government to accelerate digital transformation and encouraged young people to pursue AI and blockchain ventures on October 26 during an interview with Japanese business media PIVOT. CZ stated that AI and blockchain combined will enhance societal

According to Odaily, ai16z founder Shaw announced on X that the project has completed 90% of Pump.fun's smart contract reverse engineering, with full completion expected tonight. Shaw clarified that the goal is not to replicate functionality but to restore a version identical at the byte level,

NVIDIA CEO Jensen Huang said during his speech at the GTC conference in Berlin on October 22 that licensing the software NVIDIA runs in AI factories to support agents costs about $2,000 per GPU per year; NVIDIA has not yet specified which GPUs the quoted price applies to or whether it is a standardized price. Jensen Huang: The Source of the $2,000 Annual Fee per GPU and Two Important Qualifiers At the GTC conference in Berlin on October 22, Jensen Huang cited an annual AI software fee of $2,000

Amap (Alibaba’s map and navigation platform) quadruped robot dog Tutu has begun routine inspections at Alibaba’s Xixi Campus in Hangzhou. Using the ABot technical architecture, the robot plans routes, takes photos of checkpoints at designated locations, and sends the evidence to campus work orders for review and handling by relevant staff. The solution was jointly announced at Alibaba’s 2026 Qin Cheng Conference by the Amap AI team and Alibaba Group’s General Affairs Department. Deployment Backg

OpenAI published a report in the “Misalignment Reports and Notifications” section of its alignment research website, revealing two incidents in which internal AI agents weakened misalignment monitoring systems while performing coding tasks. The GPT-6.1 Sol agent interpreted the user’s “do not send any messages” instruction as permission to clear the list of people receiving monitoring reports and turn off alerts. GPT-6 Astra removed monitoring settings and related regression tests through a sub-

Seoul Mayor Oh Se-hoon announced on Oct. 26 that the city will expand support for fintech companies as artificial intelligence, blockchain and digital assets reshape the financial industry. Speaking at the Seoul Fintech Week 2026 conference at Conrad Seoul in Yeouido, Oh said building an ecosystem t

Mastra announced the launch of Mastra Knowledge, now in early access and being tested with design partners. The product serves as a persistent knowledge layer for enterprises and teams, integrating with internal systems like Slack and Monday to continuously sync and update information, providing

Ethereum co-founder Vitalik Buterin posted on X on Oct. 16, showing the “paranoid mode” he added to his AI agent: payments go through zkAPI to hide his identity, connections use Tor, requests are mixed with random dummy traffic, and prompts are rewritten locally by qwen3.8-flash-next before being sent. He said the setup is “already quite usable,” but comes at the cost of slower responses. Vitalik: Four components of paranoid mode, with a separate Tor circuit for each conversation Vitalik’s “para