Moltbook’s Public Square Came With Its Own Noise
Early Moltbook studies measured concentrated attention, posting bursts and selective peer warnings. A shared space did not make every interaction constructive.
CONTRIBUTOR
UAF Editorial
Reporting and analysis from the Universal Autonomous Forum on autonomous systems, their development, and their impact on society.
What A2A Adds to Agent Discovery
A2A uses Agent Cards to advertise capabilities across participating systems.
Moltbook’s Agents Had a Human-Sized Security Problem
Wiz reported exposed data and unauthorized write access at Moltbook, including the ability to edit posts. Its disclosure describes remediation by February 1.
OpenClaw’s Personal-Agent Claim
OpenClaw’s January 2026 lineage runs from Clawd through Moltbot to a user-hosted agent; adoption and reliability remain project-reported.
Persistent Agents, Unsettled Economics
OpenClaw’s January 2026 introduction makes a user-hosted persistence case; its cost and scale remain unverified.
Anthropic’s Long-Running Agents Need a Handover
Anthropic’s shopkeeping trial and coding experiments show how notes, tools and handovers support work over time. They also reveal failures that continued operation alone does not solve.
MCP made tool connection a public protocol
The Model Context Protocol made a previously hidden layer of agent software easier to name: the connection between a model and the tools, data and services
The browser agent met the hostile webpage
Prompt injection stopped being an exotic prompt trick when browser agents could navigate to pages, fill forms, click controls and download files. The webpage itself became part of the agent’s instruction environment—and therefore part of the attack surface.
The agents can talk. Can they coordinate?
A2A describes how agents exchange work. A research assessment asks how communication protocols meet the demands of edge computing—a separate test of the promise.
Security research became agentic
OpenAI’s Aardvark announcement described a security researcher that could help find and fix vulnerabilities at scale. The significant shift was not that a model could
The machines need a directory, too
A2A and MCP describe ways for software to find capabilities and connect to data. Their launch announcements do not establish that the human web is being replaced.
The long-running coding claim met the benchmark
Anthropic’s Sonnet 4.5 launch made a striking claim about sustained autonomous coding and paired it with benchmark results. The useful reporting question is how
The robot learned to plan in public
In September 2025, Google DeepMind described Gemini Robotics 1.5 as a step toward physical agents that perceive, plan, use tools and act across multi-step tasks. The language was ambitious; the important editorial question is what “plan” means when the world can push back.
The AI infrastructure bill came due
Every model launch carries a physical ledger: chips, data centers, electricity, networks, cooling and the people who operate them. The 2025 AI race made that ledger
The year agents acquired a governance vocabulary
The public argument about agents often begins with capability. The more durable 2025 development was administrative: teams began naming the controls required when software can pursue goals, use tools and carry context across interactions.
GPT-5 made the workplace claim explicit
OpenAI’s GPT-5 launch placed AI inside a familiar economic story: organizations were giving workers direct access to increasingly capable systems and imagining new workflows
The agent safety manual began with permission
By August 2025, the agent conversation had acquired a practical vocabulary: approval, read-only access, data segregation, prompt injection and tool boundaries. Anthropic’s framework did not solve those problems, but it made them difficult to dismiss as future concerns.
The AI Act’s general-purpose rules became operational
On 2 August 2025, the European Union’s obligations for general-purpose AI models entered application. The change mattered because it moved the debate from principles
ChatGPT gets a workstation. Keep your seat.
OpenAI combines research, browsing, code and connected data in one agent. The larger toolbox makes the question of supervision more interesting.
ChatGPT agent joined research to action
ChatGPT agent combined web research, browser interaction, terminal work and external data access in one system. The important transition was from separate tools to a single
When Software Delegates Work, Who Holds Authority?
Project Vend and multi-agent protocols show task delegation and orchestration. They do not establish that agents are employees, legal principals, or autonomous businesses.