Models that take actions, and what makes them reliable.
The science of agents: planning, tool use, memory, and evaluation. Posts run from ReAct and Reflexion to the benchmarks that measure whether an agent finishes a task, and what each run costs.
Oxagen 5.5 prices each model call by size before it runs, pays from a monthly allowance and then a prepaid wallet, and holds new runs when the wallet is dry while started runs keep their rules.
When a long run goes wrong, most teams can stop it or type at it. Both work badly. A steer is a third option. It is a message with a delivery mode, a status, and a record.
METR measures how long a task an agent can finish on its own. That length has doubled about every seven months since 2019. This post covers what week-long runs mean for supervision.
The worry about AI slop is about output quality. The measurements point to a different cause. Teams give an agent a workflow built for people and expect it to work.
A prompt file has no owner, no date, no scope, and no record that the agent read it. So it is a poor place for a standing rule. Steering records give each rule those things.
Without outside feedback, self-correction fails, and a model trained on its own output gets worse. This post covers what the collapse and verifier research says to do instead.
A coding agent changed your code and no one can replay how. What the research says about feedback from running code, random sampling, and turns you can audit.
Planning, tool use, memory, and reflection each come from a paper that measured something. This post traces those papers and what agents still cannot do.
An agent can answer with confidence from the wrong context. This post covers what classes, relations, constraints, and dated facts add to an agent's answers.
AgentBench, WebArena, GAIA, SWE-bench, and tau-bench each measure a different thing. None of them reports what a run costs. This post covers what that hides.
