MAROKO133 Eksklusif ai: Anthropic says it solved the long-running AI agent problem with a

📌 MAROKO133 Breaking ai: Anthropic says it solved the long-running AI agent proble

Agent memory remains a problem that enterprises want to fix, as agents forget some instructions or conversations the longer they run. 

Anthropic believes it has solved this issue for its Claude Agent SDK, developing a two-fold solution that allows an agent to work across different context windows.

“The core challenge of long-running agents is that they must work in discrete sessions, and each new session begins with no memory of what came before,” Anthropic wrote in a blog post. “Because context windows are limited, and because most complex projects cannot be completed within a single window, agents need a way to bridge the gap between coding sessions.”

Anthropic engineers proposed a two-fold approach for its Agent SDK: An initializer agent to set up the environment, and a coding agent to make incremental progress in each session and leave artifacts for the next.  

The agent memory problem

Since agents are built on foundation models, they remain constrained by the limited, although continually growing, context windows. For long-running agents, this could create a larger problem, leading the agent to forget instructions and behave abnormally while performing a task. Enhancing agent memory becomes essential for consistent, business-safe performance. 

Several methods emerged over the past year, all attempting to bridge the gap between context windows and agent memory. LangChain’s LangMem SDK, Memobase and OpenAI’s Swarm are examples of companies offering memory solutions. Research on agentic memory has also exploded recently, with proposed frameworks like Memp and the Nested Learning Paradigm from Google offering new alternatives to enhance memory. 

Many of the current memory frameworks are open source and can ideally adapt to different large language models (LLMs) powering agents. Anthropic’s approach improves its Claude Agent SDK. 

How it works

Anthropic identified that even though the Claude Agent SDK had context management capabilities and “should be possible for an agent to continue to do useful work for an arbitrarily long time,” it was not sufficient. The company said in its blog post that a model like Opus 4.5 running the Claude Agent SDK can “fall short of building a production-quality web app if it’s only given a high-level prompt, such as 'build a clone of claude.ai.'” 

The failures manifested in two patterns, Anthropic said. First, the agent tried to do too much, causing the model to run out of context in the middle. The agent then has to guess what happened and cannot pass clear instructions to the next agent. The second failure occurs later on, after some features have already been built. The agent sees progress has been made and just declares the job done. 

Anthropic researchers broke down the solution: Setting up an initial environment to lay the foundation for features and prompting each agent to make incremental progress towards a goal, while still leaving a clean slate at the end. 

This is where the two-part solution of Anthropic's agent comes in. The initializer agent sets up the environment, logging what agents have done and which files have been added. The coding agent will then ask models to make incremental progress and leave structured updates. 

“Inspiration for these practices came from knowing what effective software engineers do every day,” Anthropic said. 

The researchers said they added testing tools to the coding agent, improving its ability to identify and fix bugs that weren’t obvious from the code alone. 

Future research

Anthropic noted that its approach is “one possible set of solutions in a long-running agent harness.” However, this is just the beginning stage of what could become a wider research area for many in the AI space. 

The company said its experiments to boost long-term memory for agents haven’t shown whether a single general-purpose coding agent works best across contexts or a multi-agent structure. 

Its demo also focused on full-stack web app development, so other experiments should focus on generalizing the results across different tasks.

“It’s likely that some or all of these lessons can be applied to the types of long-running agentic tasks required in, for example, scientific research or financial modeling,” Anthropic said. 

🔗 Sumber: venturebeat.com

📌 MAROKO133 Hot ai: Long-term memory is not an ‘on/off’ switch, it’s formed by cas

Brain researchers long knew that the model for studying memory oversimplified the complex processes that the brain uses to decide what to keep and for how long. A new study demonstrated “a cascade of molecular timers unfolding across the brain regions” to store long-term memory.

For decades, memory research has focused on the hippocampus and the cortex, which are thought to be responsible for storing short-term and long-term memories. However, this model, although it did lead researchers to valuable insights, did not account for the whole picture. Why do some long-term memories last weeks while others last a lifetime?

The latest research from Priya Rajasethupathy, head of the Skoler Horbach Family Laboratory of Neural Dynamics and Cognition, builds upon previous studies, notably in 2023, that identified a pathway between short and long-term memories: the thalamus, which not only selects which memories to remember but then directs them to the cortex for long-term storage.

A press release continues that this research allowed these scientists to continue penetrating the intricate set of processes behind memory and retention.

“What happens to memories beyond short-term storage in the hippocampus? What molecular mechanisms are behind the sorting process that promotes important memories to the cortex and demotes unimportant ones to be forgotten?”

Researchers went into the lab after developing a behavioral model using a virtual reality system where mice formed specific memories. Turns out that long-term memory isn’t exactly a switch, but a rather intriguing series of “molecular timers.”

Not an on/off switch

“Existing models of memory in the brain involved transistor-like memory molecules that act as on/off switches,” says Rajasethupathy in the press release.

At the forefront of this research, the latest development in their pursuit to evolve this understanding of how the brain remembers enabled the team to “crack open this problem in a new way.”

They were able to influence how the mice remembered by varying how often certain experiences they were having were repeated. Some mice retained memories better than others. Then, they looked into the brain to investigate which mechanisms correlated with memory persistence.

Correlation was not enough, however. They needed to factor in causality. Co-lead on the study, Celine Chen, developed a CRISPR screening platform to manipulate genes in the thalamus and cortex.

“With this tool, they could demonstrate that removing certain molecules impacted the duration of the memory. Strikingly, they also observed that each molecule affected that duration on different time-scales,” as stated in a press release.

Long-term memory is an orchestration

The results then suggest that there isn’t a single on and off switch to store memories long-term, but rather there’s a “cascade of gene-regulating programs that unfold over time and across the brain like a series of molecular timers.”

“The model suggests that, after the basic memory is formed in the hippocampus, Camta1 and its targets ensure the initial persistence of the memory. With time, Tc4 and its targets are activated, providing cell adhesion and structural support to further maintain the memory. Finally, Ash1l recruits chromatin remodeling programs that make the memory more persistent.”

“Unless you promote memories onto these timers, we believe you’re primed to forget it quickly,” Rajasethupathy concludes in a press release.

The findings stand to impact memory-related brain diseases. 

Read the study in Nature.

🔗 Sumber: interestingengineering.com