An autonomy score needs a job description
OSWorld tests computer tasks. SWE-bench tests codebase repairs. Before trusting a percentage, inspect the work and the conditions behind it.
The Latest
Technology, money & power
Devin’s demo is asking for a promotion
Cognition’s coding agent comes with tools, ambition and a benchmark claim. Before handing over the repository, read the terms of the demonstration.
Your assistant remembers. Who gave it the keys?
Conversation history, application identity and delegated access solve different problems. A helpful memory should not blur them.
A saved conversation is not a standing order
OpenAI separates conversation state from execution. MemGPT manages context. Both make an assistant’s apparent memory worth examining.
In focus
The chat box is a very small window
Text prompts, learned tool calls and a Minecraft explorer reveal different machinery behind the conversation. The interface tells only part of the story.
An agent needs a finish line
ReAct, Reflexion and Toolformer offer ways to improve what an agent does next. That leaves a practical question for builders: when is enough enough?
Sometimes the clever move is to call a calculator
Toolformer learns API calls. ReAct interleaves reasoning and action. Both make a useful question concrete: what should a model check instead of guess?
More headlines
The stories behind the shifts
Attention is not a permission slip
The Transformer, GPT-3 and ReAct answer different technical questions. None of those answers should spare us from asking what a system may do.
ReAct gives the answer a next move
ReAct alternates reasoning with actions that gather information. A useful research result—and a sharper question about what we want models to do.
NEWT learned your taste in news. Sound familiar?
A news-filter prototype and ReAct reveal an older argument about software that takes initiative—and a different way to build it.
The big picture
Ideas with consequences
The helpful machine had teachers
InstructGPT learns from demonstrations and ranked answers. Before celebrating an obedient assistant, ask whose judgment helped shape the lesson.
Recognizing the object is only the beginning
Image classification and robotic grasping ask different things of vision. QT-Opt makes the action-and-feedback part explicit.
AlphaGo won the match. What else did it prove?
A victory over Lee Sedol, a documented learning system and a warning from DeepMind: keep the accomplishment and the ambition in focus.
The robot was too polite to merge
MIT’s 2007 traffic test exposed a wonderfully awkward problem: a driverless car could be too cautious. DARPA made autonomy answer to a road course.
Before OpenClaw, the Lobster Changed Its Name
The project’s own history traces several names. It offers a lineage, not an independent verdict on adoption or autonomy.
What Moltbook Actually Demonstrated
Moltbook presented itself as a social network for AI agents, but human verification and scheduled automation complicate claims of machine independence. An arXiv analysis measured shallow interaction.
AutoGPT and an Early Delegation Experiment
AutoGPT made goal-driven delegation visible, then exposed the costs of loops, hallucinations, and unreliable workflows.
AI Agents Became a Product Category, With Caveats
Retrospectives disagree about the agent boom. Product launches are visible; reliable deployment and comparable adoption measures remain harder to establish.
Who Profits When the Machines Take Over the Work? Tracing Value in the Age of Software Delegation
Who captures value when software takes on more work? The evidence points to data, workflow knowledge, and human accountability, while leaving the distribution of profits unsettled.
The Agent Can Act. Who Answers for It?
Economic participation raises a harder question than task completion: whose preferences guide the agent, and what happens when its choices go wrong?
The world’s factories now operate five million industrial robots
The International Federation of Robotics estimates that the global stock of industrial robots in operation reached five million units in 2025. New installations rose to more than 600,000, an annual flow that adds to—but is smaller than—the installed base.
AGIBOT says more than 300 robots are entering a Chinese theme park
Chinese robotics maker AGIBOT says it and Chimelong have begun a large-scale deployment involving more than 300 robots across entertainment, education, visitor services and hotel operations. The announcement describes an ambitious live setting; it does not yet offer independent evidence of how…
A robot that can move still needs a worksite
Robotics coverage often begins with the body: can the machine walk, grasp, see or navigate? Deployment begins somewhere less cinematic. It begins with a worksite, a task boundary, a maintenance plan and a person who knows when to stop the machine.
Senators propose pre-release checks and incident reporting for frontier AI
A bill introduced by U.S. Senators Mark Warner and Brian Schatz would create a federal AI Safety Board, require pre-release access to certain frontier models, and set reporting rules for serious incidents. It is a proposal; none of its requirements are currently in force.
Delegated Access Leaves Questions for Operators
SPIFFE’s Workload Endpoint omits direct client authentication and relies on out-of-band authenticity checks; OAuth tokens represent scoped, time-limited authorization.
The Autonomous World Needs a Public Record
NIST offers a voluntary framework for AI risk. California’s DMV points the public to autonomous-vehicle incident reporting and collision records. These are different tools for asking what happened—and what deserves scrutiny.
Keep reading
From the UAF archive
Your clock has a purpose. Does its behavior?
Being built for a purpose is different from acting toward a goal. Two accounts of machine behavior give that distinction teeth.