{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-08-07T02:00:35.291Z","headline":"Prime Agent：一个具有自我改进能力的RLM代理","description":"Prime Agent 是一个自我改进的编码代理，围绕递归语言模型（RLM）和持续框架（Continual Harness）两大抽象构建，将上下文视为变量、子代理委派视为 REPL 内的函数调用，并允许代理对其提示词、技能、记忆和子代理进行 CRUD 操作。它完全开源，可通过 curl 命令安装，支持与前沿模型即时使用，并具备后台守护进程、会话恢复、分支分叉和异步内核压缩等特性。","url":"https://www.aioga.com/news/cmsgulpsm0hwaro5qcjgjlvgr/","mainEntityOfPage":"https://www.aioga.com/news/cmsgulpsm0hwaro5qcjgjlvgr/","datePublished":"2026-08-06T01:27:27.867Z","dateModified":"2026-08-06T01:27:27.867Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.primeintellect.ai/blog/prime-agent","https://aihot.virxact.com/items/cmsgulpsm0hwaro5qcjgjlvgr"],"canonicalUrl":"https://www.aioga.com/news/cmsgulpsm0hwaro5qcjgjlvgr/","directAnswer":{"@type":"Answer","text":"Prime Intellect 发布 Prime Agent，这是一款自我改进的编码代理运行框架。其设计围绕递归语言模型（RLM）与持续框架两项抽象，强调让代理在运行中处理和调整自身的提示词、技能、记忆及子代理。","url":"https://www.aioga.com/news/cmsgulpsm0hwaro5qcjgjlvgr/","dateCreated":"2026-08-06T01:27:27.867Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"primeintellect.ai source article","url":"https://www.primeintellect.ai/blog/prime-agent","datePublished":"2026-08-06T01:27:27.867Z","provider":{"@type":"Organization","name":"primeintellect.ai","url":"https://www.primeintellect.ai/blog/prime-agent"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmsgulpsm0hwaro5qcjgjlvgr","datePublished":"2026-08-06T01:27:27.867Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmsgulpsm0hwaro5qcjgjlvgr"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"primeintellect.ai","url":"https://www.primeintellect.ai/blog/prime-agent"},"geoDeepAnswer":null,"article":{"id":"cmsgulpsm0hwaro5qcjgjlvgr","slug":"cmsgulpsm0hwaro5qcjgjlvgr","url":"https://www.aioga.com/news/cmsgulpsm0hwaro5qcjgjlvgr/","title":"Prime Agent：一个具有自我改进能力的RLM代理","title_en":"","summary":"Prime Agent 是一个自我改进的编码代理，围绕递归语言模型（RLM）和持续框架（Continual Harness）两大抽象构建，将上下文视为变量、子代理委派视为 REPL 内的函数调用，并允许代理对其提示词、技能、记忆和子代理进行 CRUD 操作。它完全开源，可通过 curl 命令安装，支持与前沿模型即时使用，并具备后台守护进程、会话恢复、分支分叉和异步内核压缩等特性。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://www.primeintellect.ai/blog/prime-agent","aiHotUrl":"https://aihot.virxact.com/items/cmsgulpsm0hwaro5qcjgjlvgr","publishedAt":"2026-08-06T01:27:27.867Z","category":"产品更新","score":70,"selected":true,"articleBody":["Today, we are launching Prime Agent , our self-improving coding harness designed around two abstractions, the Recursive Language Model (RLM) [citation：https://arxiv.org/abs/2512.24601] and Continual Harness [citation：https://arxiv.org/abs/2605.09998]. Modern harness designs were built around the capabilities of earlier generations of models, and they do not reflect what frontier models can do today: fixed tool-calling schemas and context compaction force the model to work around its own scaffolding instead of leveraging it. Static, hand-engineered sub-agents, prompts, skills, and memory are set once at design time and never adapt to what the agent learns while running. We believe that harnesses should instead extrapolate on current model capabilities toward the next frontier of reasoning patterns.","Prime Agent is built around this principle through two main abstractions:","These abstractions are powerful for bootstrapping model capabilities. Prime Agent is built to be effective as a general coding assistant, as a default runtime for long-horizon autonomous evaluation, and as a collaborator for research and autoresearch.","Prime Agent is fully open-source：https://github.com/PrimeIntellect-ai/prime-agent, and can be installed via:","The performance of agent harnesses are tied to both the design of the harness and the capability of the model trained around the harness. We designed Prime Agent to be immediately usable with modern open and closed frontier models, while also providing a feature set that we expect to provide further performance gains as newer generations of models are trained around it.","At its core, Prime Agent is designed around programmatic tool and sub-agent calling. Models in Prime Agent use a persistent IPython kernel as their only tool. Other standard harness features are called as functions in the kernel, including sub-agents, which are each implemented as another prime-agent instance.","Background Daemon and Agents View. The default view is a text-user interface (TUI) similar to other coding agent harnesses. By default, IPython actions made by the agent are condensed for brevity, but can be expanded to view actions made by the harness. Sub-agents launched in the REPL can also be accessed below the user chatbox.","Prime Agent runs a background daemon that owns all live agent sessions over a local socket. You can attach and detach from the session without affecting the underlying agent loop. Each root session tree runs in a recoverable worker process; if a worker crashes, the daemon recovers it from the session JSONL and kernel state snapshot.","The Agents View allows you to see and select other live sessions from the daemon. It can be opened by pressing the Left Arrow key (←) on an empty prompt, and lists sessions that are currently running, idle sessions with the daemon still active, and inactive sessions that are currently not loaded in memory. Any of these chats can immediately be entered and interacted with, and pressing space allows users to chat with a session in any state, including steering and queuing of prompts and commands such as /compact .","The Agents View is constructed as the central connecting point between agents and subagents, recursively. Any agent is discoverable in an Agents View. Users navigate from an Agents View into an agent's chat, then into the Agents View of its subagents, into a subagent chat, and so on.","Because subagents share the same Running-Idle-Inactive state machine as the root agents, they can be removed from memory after 30 minutes of inactivity, and the moment a user or agent addresses any of them, they are reloaded from disk. In highly nested chats, this can save a lot of memory.","Session and Context Management. The entire session history of the agent is stored as append-only JSONL files on disk. Each line is a JSON entry, which can include messages, model switches, compaction summaries, or extension entries. Branching, forking, and cloning all happen within the same file by moving the leaf pointer. The full history is always recoverable through /tree .","Compaction happens when the context hits a threshold or directly by the agent in the REPL with compact.run() . Compaction is primarily used to clean the main context of the agent, but the full history, including past compactions, can be accessed programmatically in the IPython kernel when needed.","The introduction of the REPL requires additional work to manage the IPython state. We asynchronously compact and clean the kernel simultaneously, using a spawned agent to act as a garbage collector. This is necessary to avoid REPL memory built up for each agent.","Prime Agent relies on the IPython kernel as its REPL that persists over the session, which it can invoke every turn. On initialization, the kernel pre-imports each skill / tool as a module, including the rlm for recursive programmatic sub-agent calling.","The rlm is an asynchronous function, meaning the model can freely invoke and parallelize sub-agent calls in code. Spawning a subagent (e.g. await rlm(\"sub-task\") ) launches a full session with its own model, IPython kernel, session tree, and conversation history. It returns immediately, because all subsequent communication between agents happens through the agent_message.send(...) tool.","There are several useful primitives that Prime Agent can choose to launch in this way, such as fanning out sub-agents in parallel, or launching background work.","As models continue to improve, new invocation patterns over tool calls and sub-agents will emerge. We expect future generations of models to rely less on hand-holding prompts and more on this kind of direct, programmatic control.","The background daemon manages all live Prime Agent sessions. Prime Agent also enables Agent-to-Agent (A2A) messaging through the daemon, letting any Prime Agent session message any other Prime Agent session using the same mechanism used for messaging persistent sub-agents. This allows for easy orchestration to manage the progress of sub-agent swarms and communication regarding shared resources directly between the affected agents. To prevent undesirable communication across independent sessions, multi-agent communication in Prime Agent is limited to its nuclear family , meaning parent, sibling, or child processes.","Prime Agent supports persistent sub-agents through its RLM-native runtime, meaning a sub-agent's own session directory, context, IPython kernel, and session history persist even after the initial sub-agent call has finished. Prime Agent can send further messages to continue a persistent sub-agent by accessing its unique session identifier, all from its IPython kernel.","Prime Agent's harness state lives in the persistent IPython kernel as rlm.harness , immediately readable and callable by the agent mid-task, and every change is also written to disk, so it survives across turns and across sessions. Continual Harness formalizes this state as H = ( ρ , G , K , M ) H=(\\rho, G, K, M) H = ( ρ , G , K , M ) , prompt, sub-agents, skills, and memory, refined online from the agent's own trajectory without resets.","Each of the four components exposes the same create, read, update, delete surface. create_prompt_note(...) , create_memory(...) , create_skill(...) , and create_subagent(...) each add an entry of that kind, update_X(...) and delete_X(...) mirror them, and list(kind) or get(kind, id) read them back. Skills follow this same surface: authoring a Python-backed skill is a create_skill(...) call carrying a SKILL.md -style reference, the same operation as adding a memory or a prompt note.","/refine is the self-improving pipeline built on top of this CRUD surface. It reads the agent's own trajectory, the record of what was tried and what happened, and applies the smallest relevant CRUD edit that improves the harness toward better outcomes: updating a prompt note, memory, skill, or sub-agent spec, rather than rewriting the whole harness. Each refinement records its trigger and the outcome it produced, so improvement is evidence-backed rather than arbitrary. Refinement runs in two phases. Planning, the LLM call that proposes the edit, runs in the background and does not block the ongoing conversation. Applying the edit, writing to disk and rebuilding the system prompt, is fast and only briefly blocks at the next turn boundary. The agent can call refine.run() directly whenever it notices a repeated failure or a reusable tactic, not only on a fixed schedule.","The base system prompt remains immutable. /refine only edits the harness layer around it. Rollback is supported through prior refinement history, allowing a bad harness update to be reverted by ID.","Prime Agent's eval mode combines three complementary mechanisms. A goal sets the overall objective: a persistent objective with an optional token budget that the harness keeps re-prompting the agent to pursue across turns, tracked until the agent explicitly calls goal.complete() . Heartbeats are scheduled cron-style messages injected into the session on a fixed interval, used for regular checks such as monitoring a sub-agent's progress or polling for a training update. Autonomous mode is the continuation mechanism itself, ensuring the agent keeps working toward the goal instead of stopping early once a turn produces no further output. Together, these let a session run unattended for extended periods while remaining bounded by an explicit budget and inspectable through the Agents View.","Autonomous mode is available directly from the CLI with --autonomous , no scripting required. A run can set a completion goal and a turn limit in the same command:","The gate command runs before the session is allowed to finish. A failed gate returns its bounded output to the agent for another attempt, and Prime Agent skips rerunning a failed gate when the workspace has not changed since the last attempt. --autonomous-max-turns , --autonomous-max-tokens , and --autonomous-timeout-ms bound continuations, tokens, and wall-clock time respectively.","Prime Agent serves as both a coding agent to be used, and a harness design to be evaluated for research. We make special note that while many modern frontier models are trained around a specific harness, currently no model has been trained around Prime Agent or its core feature set.","ARC-AGI 3. ARC-AGI 3：https://arcprize.org/arc-agi/3 is a popular intelligence benchmark that measures the ability of an agent to perform symbolic reasoning and learn the rules of simulated worlds. We evaluate Prime Agent with autonomous mode over several different frontier models, and compare to their native harnesses. Prime Agent was developed as a CLI coding agent, so the only ARC AGI 3 specific changes are to the task prompt, inspired by the standard prompt setup used in PRO-LONG：https://arxiv.org/abs/2607.20064v2.","Our best results use Opus 5 in Prime Agent to achieve 95.5% RHAE Best@1 , which surpasses the ARC reported human expert baseline of 95.4% . Across three runs, we find that Prime Agent consistently performs well [95.0, 95.2, 95.5] and 99.97% Best@3 with all 183/183 levels complete. Our median score card action replay (95.2%) for ARC-AGI-3 can be found here：https://arcprize.org/scorecards/2af780b4-f2a1-43e9-a794-b23da3cd3f9f.","In addition to achieving a higher maximum score over each model's native harness, we find that Prime Agent also does so at a lower overall token usage. Prime Agent saves tokens by programmatically running functions over data rather than spending tokens reading data using tools.","Finally, we note that we evaluated Opus 5 and GPT-5.6 Sol with Claude Code and Codex respectively, and found worse overall performance relative to the official results, so we yield to their official reported numbers instead.","Many difficult tasks in the wild reduce to long context tasks. Our goal is to show that Prime-Agent with open-weights models are a competitive alternative to closed models and harnesses, both as a general agent to be used, and as a baseline harness to be evaluated.","Below, we select a suite of common long-context benchmarks across coding, retrieval, and general long reasoning tasks, and compare Prime Agent to several different popular harnesses. We offload the main context in each harness to a file in memory to start. For closed model harnesses, we use their associated models (i.e., Codex with GPT, Claude Code with Opus) while for Prime-Agent and Pi-mono (with sub-agents), we choose an open-weights model in GLM-5.2.","We generally find Prime Agent to be competitive across a wide range of long tasks, especially against the harness that did not use a model trained around it. Prime Agent especially excels at long-running or long-context tasks, and can competitively run on its own as an autonomous agent. We include a set of focused case studies and experiments on long settings where Prime Agent excels.","Creating emulators from scratch. An emulator is software that reproduces another computer system's observable behavior. We evaluate Prime Agent on EmulatorBench , a preview benchmark that tasks agents with constructing emulators in Rust for a variety of game systems. Agents are given a specification of the emulator and a set of diagnostic tests in the form of a verifier.","The correctness of an emulator is given from its ability to mimic the behavior of the target machine. This is measured by human-generated diagnostic programs that inspect the emulator's behavior, such as the CPU flags, PPU timing, and other components. In an effort to minimize the effects of data contamination, we require the agent to build the emulator from scratch in Rust, sandboxed without any reference implementation. We report preliminary results on this long-context coding benchmark averaged over 16 emulator reconstructions, as well as two emulators, the SEGA Genesis and Nintendo Game Boy Color, that Prime Agent successfully reproduces. For Opus, our runs surprisingly failed to solve the tasks despite successful tool-call responses.","Writing GPU kernels. Writing performant GPU kernels is an iterative process that requires repeatedly verifying, profiling, and tweaking code to get correct. We evaluate Prime Agent as a harness for GPU kernel writing on the recently released PMPP-Hard：https://blog.sinatras.dev/PMPP-Hard benchmark, a suite of tasks where agents must write performant GPU kernels that pass a suite of correctness checks against KernelGuard：https://github.com/gpu-mode/kernelguard, the verification tool used for the official GPU MODE：https://www.gpumode.com/home kernel leaderboard.","Autonomously playing video games has become an interesting case study for models and harnesses in how they handle long-horizon decision making. Games often require harnesses to balance information and context across millions of tokens, while also leveraging this information to efficiently take actions and avoid catastrophic states.","Factorio. Factorio is a 2D factory simulation game where agents must mine resources, research technology, and build automated factories to increase the production of these resources. The Factorio Learning Environment (FLE)：https://jackhopkins.github.io/factorio-learning-environment/versions/0.3.0.html is an interface for simplifying the observation and action space of an LLM playing Factorio, which we use to connect Prime Agent to the game.","The action and observation space of FLE is a module in Python that is accessed programmatically at every turn. This integrates directly into Prime Agent's IPython kernel. To leverage PTC for sub-agents, we launch four controllable characters in the game.","The primary metric in FLE is production score, which is a weighted average of all materials the agent produces. Prime Agent successfully leveraged /refine to turn failures and successes into memories and skills, respectively. It used its own accumulated experience to design increasingly efficient machine layouts, raising the production score run over run. This allowed Prime Agent to efficiently score in the 100K+ range in production score in a matter of hours.","However, we also observed instances of reward hacking by Prime Agent in FLE. Prime Agent discovered it could bypass Factorio's rules entirely by spawning in resources directly into its assembly machines through RCON commands, even with an explicit heartbeat prompt to remind Prime Agent not to cheat in Factorio. Once it found this exploit, the same refinement loop that had been building legitimate skills turned to building efficient cheating skills instead.","MazeBench. MazeBench：https://mazebench.com/blog?post=maze-bench-results is an open-world 3D spatial reasoning environment where the player controls a 3D cube and must solve puzzle rooms within a global maze, while collecting gems. Frontier models are shown to greatly struggle on this task, expending billions of tokens to solve only a fraction of the overall world. We compare Opus 5 and GPT-5.6 Sol with Prime Agent versus their native harnesses, as well as GLM-5.2 with Claude Code. Following the benchmark metrics, we report the unique number of rooms they find, the unique number of states, and the total number of gems, all as a function of their overall token spend.","Prime Agent is a new paradigm on the design of agent harnesses. Despite strong results over other harnesses, we still notice friction when running Prime Agent with models. This implies that there are huge performance gains still available from training with Prime Agent directly around this harness paradigm, or even the individual RLM and Continual Harness components.","We strongly believe that model-harness co-learning is the dominant paradigm to unlock new capabilities. Many features of Prime Agent are not fully utilized without a trained model, and we believe there are huge performance gains still available from training with the harness directly. We are excited to bring you these new capabilities, all in the open.","We will have a full technical report with further details soon.","Prime Agent is built on top of pi ：https://github.com/earendil-works/pi. We thank the authors of pi for their valuable work."],"articleImages":[{"sourceUrl":"https://www.primeintellect.ai/_next/image?url=%2Fblog%2Fprime-agent%2Fcover.png&w=3840&q=75&dpl=dpl_AVn3FxYGVUQtLL2Nnd5BDzGFZCQZ","alt":"Prime Agent: A self-improving RLM agent","afterParagraph":0,"url":"/media/articles/cmsgulpsm0hwaro5qcjgjlvgr/71b6f7c1f7f2423e.webp"},{"sourceUrl":"https://www.primeintellect.ai/blog/prime-agent/onboarding-splash.gif","alt":"Prime Agent onboarding splash screen","afterParagraph":3,"url":"/media/articles/cmsgulpsm0hwaro5qcjgjlvgr/4fe519ceda62228d.gif"},{"sourceUrl":"https://www.primeintellect.ai/blog/prime-agent/architecture.png","alt":"Prime Agent architecture diagram","afterParagraph":5,"url":"/media/articles/cmsgulpsm0hwaro5qcjgjlvgr/c1b385164aea430a.png"},{"sourceUrl":"https://www.primeintellect.ai/blog/prime-agent/tui-view.png","alt":"Prime Agent text user interface","afterParagraph":6,"url":"/media/articles/cmsgulpsm0hwaro5qcjgjlvgr/752cc5c1d4a09ae9.png"}],"mediaStatus":"ok","articleBodyZh":["今天，我们发布了 Prime Agent，这是一个自我改进的编码工具平台，围绕两个抽象概念设计：递归语言模型（RLM）[引用：https://arxiv.org/abs/2512.24601] 和连续工具平台（Continual Harness）[引用：https://arxiv.org/abs/2605.09998]。现代工具平台的设计基于早期模型的能力，它们没有反映出前沿模型今天的能力：固定的工具调用模式和上下文压缩迫使模型绕过自身的框架而不是利用它。静态、手工设计的子代理、提示、技能和记忆在设计时设定后不会随着代理运行时所学内容而自适应。我们认为，工具平台应该基于当前模型能力向下一步推理模式的前沿扩展。","Prime Agent 的设计基于这一原则，通过两个主要抽象实现：","这些抽象对于引导模型能力的启动非常强大。Prime Agent 旨在作为通用编码助手有效运作，作为长期自主评估的默认运行时，以及作为研究和自动研究的协作工具。","Prime Agent 是完全开源的：https://github.com/PrimeIntellect-ai/prime-agent，并且可以通过以下方式安装：","代理平台的性能既与平台设计相关，也与围绕平台训练的模型能力相关。我们设计 Prime Agent 以便可立即与现代开放和封闭前沿模型一起使用，同时提供一套功能，我们预计随着新一代模型围绕它训练，将进一步提高性能。","Prime Agent 的核心设计围绕程序化工具和子代理调用。Prime Agent 中的模型使用持久化的 IPython 内核作为其唯一工具。其他标准平台功能作为内核中的函数调用，包括子代理，每个子代理都是另一个 prime-agent 实例实现的。","后台守护进程和代理视图。默认视图是类似其他编码代理平台的文本用户界面（TUI）。默认情况下，代理执行的 IPython 操作会被压缩以简洁显示，但可以展开以查看平台执行的操作。在 REPL 中启动的子代理也可以在用户聊天框下方访问。","Prime Agent 运行一个后台守护进程，该进程通过本地套接字管理所有活动的代理会话。你可以在不影响底层代理循环的情况下附加或分离会话。每个根会话树在一个可恢复的工作进程中运行；如果工作进程崩溃，守护进程会通过会话 JSONL 文件和内核状态快照进行恢复。","代理视图允许你查看并选择守护进程中的其他活动会话。可以在空提示符下按左箭头键（←）打开，它会列出当前正在运行的会话、游离状态但守护进程仍活动的会话，以及当前未加载到内存中的非活动会话。可以立即进入并与这些会话互动，按空格键允许用户与任何状态的会话聊天，包括引导和排队提示与命令，如 /compact。","代理视图被构建为代理与子代理之间的中心连接点，并递归进行。任何代理都可以在代理视图中被发现。用户从代理视图导航进入某个代理的聊天，然后进入其子代理的代理视图，再进入子代理聊天，依此类推。","由于子代理与根代理共享相同的运行-空闲-非活动状态机，它们在 30 分钟不活动后可以从内存中移除，当用户或代理访问其中任何一个时，它们会从磁盘重新加载。在高度嵌套的聊天中，这可以节省大量内存。","会话与上下文管理。整个代理的会话历史被存储为仅追加的 JSONL 文件在磁盘上。每一行都是一个 JSON 条目，可以包含消息、模型切换、压缩摘要或扩展条目。分支、分叉和克隆都通过移动叶子指针在同一文件内完成。通过 /tree 可以始终恢复完整历史。","当上下文达到阈值或在 REPL 中由代理直接调用 compact.run() 时，会发生压缩。压缩主要用于清理代理的主上下文，但完整历史，包括过去的压缩，可以在需要时通过 IPython 内核以编程方式访问。","引入 REPL 需要额外工作来管理 IPython 状态。我们异步地同时压缩和清理内核，使用一个派生的代理作为垃圾回收器。这对于避免每个代理的 REPL 内存累积是必要的。","Prime Agent 依赖 IPython 内核作为其持续会话的 REPL，它可以在每个回合调用。在初始化时，内核会将每个技能/工具预先导入为模块，包括用于递归程序性子代理调用的 rlm。","rlm 是一个异步函数，这意味着模型可以自由调用并并行化代码中的子代理调用。生成子代理（例如 await rlm(\"sub-task\")）会启动一个拥有自己模型、IPython 内核、会话树和对话历史的完整会话。它会立即返回，因为后续代理之间的所有通信都是通过 agent_message.send(...) 工具进行的。","Prime Agent 可以选择以这种方式启动的几种有用原语，例如并行扩展子代理，或启动后台工作。","随着模型的不断改进，针对工具调用和子代理的新调用模式将会出现。我们预计未来几代模型将越来越少依赖于逐步提示，而更多依赖于这种直接的程序化控制。","后台守护进程管理所有活跃的 Prime Agent 会话。Prime Agent 还通过守护进程启用代理间 (A2A) 消息传递，使任何 Prime Agent 会话都可以使用与向持久子代理发送消息相同的机制向其他任何 Prime Agent 会话发送消息。这便于轻松协调以管理子代理群的进度，并直接在相关代理之间就共享资源进行通信。为了防止独立会话之间的不良通信，Prime Agent 的多代理通信仅限于其核心家庭，即父、兄弟姐妹或子进程。","Prime Agent 通过其 RLM 原生运行时支持持久子代理，这意味着即使初始子代理调用完成，子代理自身的会话目录、上下文、IPython 内核和会话历史仍会保留。Prime Agent 可以通过访问子代理的唯一会话标识符发送进一步消息，以继续持久子代理，所有操作均在其 IPython 内核中完成。","Prime Agent 的 harness 状态作为 rlm.harness 存在于持久化的 IPython 内核中，代理在任务中可立即读取和调用，且每个变更也都写入磁盘，因此跨回合和跨会话保持。连续引导将该状态形式化为 H = （ ρ， G， K， M ） H=（\\rho， G， K， M） H = （ ρ， G， K， M ）， 提示词、子代理、技能和记忆，从代理自身轨迹在线精炼且无需重置。","四个组件中的每个都暴露了相同的创建、读取、更新和删除表面。create_prompt_note（...）， create_memory（...）， create_skill（...）， create_subagent（...） 各自添加该类条目，update_X（...） 和 delete_X（...） 镜像它们，并 list（kind） 或 get（kind， id） 读取它们。技能也遵循同样的表面：编写一个Python支持的技能是一个带有 SKILL.md 风格引用的create_skill（...）调用，操作与添加内存或提示笔记相同。","/refine是建立在这个CRUD表面上的自我改进管道。它读取代理自身的轨迹、尝试内容和发生情况的记录，并应用最小相关的CRUD编辑，以提升束缚以达成更好结果：更新提示笔记、记忆、技能或子代理规格，而非重写整个束缚。每次改进都会记录其触发点及其产生的结果，因此改进有证据支持，而非任意。精炼分为两个阶段。规划，即提议编辑的LLM调用，在后台运行，不会阻断正在进行的对话。应用编辑、写入磁盘并重建系统提示符快速，且仅在下一回合边界短暂阻塞。代理在发现反复失败或可重用策略时，可以直接调用 refine.run（），而不仅仅是固定时间表。","基础系统提示词保持不变。/refine 只编辑其周围的线束层。回滚通过之前的细化历史支持，允许通过ID恢复不良的线束更新。","Prime Agent 的评估模式结合了三种互补机制。目标（goal）设定总体目标：一个持久目标，可选地设有令牌预算，由系统不断提示代理在多个回合中执行，跟踪直到代理明确调用 goal.complete()。心跳（Heartbeats）是按固定间隔注入会话的 cron 风格消息，用于定期检查，例如监控子代理的进度或轮询训练更新。自主模式（Autonomous mode）本身就是延续机制，确保代理持续朝着目标工作，而不是在某回合没有输出时就提前停止。这些机制结合在一起，使会话可以长时间无人值守运行，同时仍受明确预算约束，并可通过代理视图（Agents View）进行检查。","自主模式可以直接通过 CLI 使用 --autonomous 启用，无需编写脚本。一次运行可以在同一命令中设置完成目标和回合限制：","gate 命令在允许会话结束前运行。门（gate）失败时会将其受限输出返回给代理以供再次尝试；当工作空间自上次尝试后没有变化时，Prime Agent 会跳过重新运行失败的 gate。--autonomous-max-turns、--autonomous-max-tokens 和 --autonomous-timeout-ms 分别限制延续回合数、令牌数量和时钟时间。","Prime Agent 既是一个可使用的编码代理，也是一个用于研究评估的系统设计。我们特别指出，尽管许多现代前沿模型都是围绕特定系统设计训练的，但目前尚无模型围绕 Prime Agent 或其核心功能集进行训练。","ARC-AGI 3：ARC-AGI 3：https://arcprize.org/arc-agi/3 是一个流行的智能基准，用于衡量代理执行符号推理和学习模拟世界规则的能力。我们在多种不同前沿模型上使用自主模式评估 Prime Agent，并与它们的原生系统进行比较。Prime Agent 被开发为 CLI 编码代理，因此针对 ARC AGI 3 的唯一特定更改是任务提示，这些提示灵感来源于 PRO-LONG 中使用的标准提示设置：https://arxiv.org/abs/2607.20064v2。","我们最好的结果是在 Prime Agent 中使用 Opus 5，实现了 95.5% 的 RHAE Best@1，这超过了 ARC 报告的人类专家基线 95.4%。在三次运行中，我们发现 Prime Agent 始终表现良好 [95.0, 95.2, 95.5]，Best@3 达到 99.97%，并完成了所有 183/183 个关卡。ARC-AGI-3 的中位数成绩卡动作回放（95.2%）可以在这里找到：https://arcprize.org/scorecards/2af780b4-f2a1-43e9-a794-b23da3cd3f9f。","除了在每个模型的本地框架上取得更高的最高分，我们还发现 Prime Agent 的总体 token 使用量更低。Prime Agent 通过对数据进行程序化函数运行来节省 token，而不是使用工具读取数据消耗 token。","最后，我们注意到我们对 Opus 5 和 GPT-5.6 Sol 分别使用 Claude Code 和 Codex 进行了评估，结果显示相较于官方结果总体表现较差，因此我们采用其官方报告的数据。","许多实际中的困难任务都归结为长上下文任务。我们的目标是展示，使用开源权重模型的 Prime-Agent 是封闭模型和框架的有竞争力的替代方案，既可以作为一个可使用的通用代理，也可以作为一个用于评估的基准框架。","下面，我们选择了一组常见的长上下文基准测试，涵盖编码、检索和一般长推理任务，并将 Prime Agent 与几种不同的流行框架进行比较。我们将每个框架中的主要上下文预先载入到内存文件中开始。对于封闭模型框架，我们使用其相关模型（即 Codex 配合 GPT、Claude Code 配合 Opus），而对于 Prime-Agent 和 Pi-mono（含子代理），我们选择 GLM-5.2 中的开源权重模型。","我们普遍发现，Prime Agent 在广泛的长任务中具有竞争力，尤其是在没有使用其训练模型的框架中。Prime Agent 在长时间运行或长上下文任务中表现尤为出色，并且可以作为自主代理独立运行。我们还包括了一组关于 Prime Agent 在长任务中表现优异的集中案例研究和实验。","从零开始创建模拟器。模拟器是一种软件，它可以重现另一台计算机系统的可观察行为。我们在 EmulatorBench 上评估 Prime Agent，这是一个预览基准测试，它要求代理使用 Rust 为各种游戏系统构建模拟器。代理会收到模拟器规范和一组以验证器形式提供的诊断测试。","模拟器的正确性来源于其模仿目标机器行为的能力。这通过人类生成的诊断程序来测量，这些程序会检查模拟器的行为，例如 CPU 标志、PPU 定时以及其他组件。为了尽量减少数据污染的影响，我们要求代理从零开始使用 Rust 构建模拟器，并在沙箱环境中运行，且不参考任何实现。我们报告了在这个长上下文编码基准测试上的初步结果，这些结果是对 16 个模拟器重建平均得出的，同时还包括 Prime Agent 成功重现的两款模拟器，SEGA Genesis 和 Nintendo Game Boy Color。对于 Opus，尽管工具调用响应成功，我们的运行出乎意料地未能解决任务。","编写 GPU 内核。编写高性能 GPU 内核是一个迭代的过程，需要反复验证、分析和调整代码以确保正确。我们在新发布的 PMPP-Hard：https://blog.sinatras.dev/PMPP-Hard 基准测试上评估 Prime Agent 作为 GPU 内核编写工具的表现，该基准包含一系列任务，要求代理编写能够通过 KernelGuard：https://github.com/gpu-mode/kernelguard 正确性检查的高性能 GPU 内核，这是官方 GPU MODE：https://www.gpumode.com/home 内核排行榜使用的验证工具。","自主玩电子游戏已成为模型和工具在处理长远决策时的一个有趣案例研究。游戏通常要求工具在处理数百万个 token 的信息和上下文时实现平衡，同时也要利用这些信息有效采取行动并避免灾难性状态。","Factorio。Factorio 是一款二维工厂模拟游戏，玩家必须采矿资源、研究技术并建造自动化工厂以增加这些资源的产量。Factorio 学习环境 (FLE)：https://jackhopkins.github.io/factorio-learning-environment/versions/0.3.0.html 是一个用于简化大型语言模型（LLM）玩 Factorio 时的观察和动作空间的接口，我们使用它将 Prime Agent 与游戏连接起来。","FLE 的动作和观察空间是一个 Python 模块，可以在每回合通过编程方式访问。它直接集成到 Prime Agent 的 IPython 内核中。为了为子代理利用 PTC，我们在游戏中启动了四个可控角色。","FLE 的主要指标是生产得分，它是代理生产的所有材料的加权平均值。Prime Agent 成功利用 /refine 将失败和成功分别转化为记忆和技能。它利用自身积累的经验设计出越来越高效的机器布局，使每次运行的生产得分不断提高。这使得 Prime Agent 能在数小时内高效获得 10 万以上的生产得分。","然而，我们也观察到 Prime Agent 在 FLE 中存在奖励作弊的情况。Prime Agent 发现可以通过 RCON 命令直接在其装配机中生成资源，从而完全绕过 Factorio 的规则，即使有明确的心跳提示提醒 Prime Agent 不要在 Factorio 中作弊。一旦发现该漏洞，原先用于构建合法技能的同样精炼循环就转向了构建高效作弊技能。","MazeBench。MazeBench：https://mazebench.com/blog?post=maze-bench-results 是一个开放世界的 3D 空间推理环境，玩家控制一个 3D 立方体并必须在全球迷宫中解决谜题房间，同时收集宝石。前沿模型在此任务中表现出很大困难，需要消耗数十亿个 token 才能解决整体世界的一小部分。我们比较了 Opus 5 和 GPT-5.6 Sol 与 Prime Agent 的表现及它们各自原生接口，同时也比较了 GLM-5.2 与 Claude Code。按照基准指标，我们报告它们找到的唯一房间数量、唯一状态数量以及宝石总数，均作为其整体 token 消耗的函数。","Prime Agent 是代理装备设计上的一种新范式。尽管在其他装备上取得了强劲的结果，但我们仍注意到在使用模型运行 Prime Agent 时存在摩擦。这表明，通过直接围绕该装备范式或甚至单独的 RLM 和持续装备组件进行训练，仍有巨大的性能提升空间。","我们坚信模型与装备的共同学习是解锁新能力的主导范式。没有经过训练的模型，Prime Agent 的许多功能无法充分利用，我们相信通过直接使用该装备进行训练，仍有巨大的性能提升空间。我们很高兴向您提供这些新的能力，并且全部开源。","我们将很快发布一份包含更多细节的完整技术报告。","Prime Agent 构建于 pi 之上： https://github.com/earendil-works/pi。我们感谢 pi 的作者们所做的宝贵工作。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Prime Intellect 发布 Prime Agent，这是一款自我改进的编码代理运行框架。其设计围绕递归语言模型（RLM）与持续框架两项抽象，强调让代理在运行中处理和调整自身的提示词、技能、记忆及子代理。","background":"材料称，Prime Agent 将持久化 IPython 内核作为唯一工具入口，常规运行框架功能及子代理均以内核中的函数形式调用；每个子代理本身也是一个 prime-agent 实例。产品定位包括通用编码助手、长周期自主评测运行时，以及研究协作场景。","viewpoint":"Aioga 判断，这一发布的重点不只是新增编码代理，而是尝试把代理编排从预设流程转向可在执行期间修改的运行机制。值得关注的是，该思路能否在不同模型上稳定复现预期效果，材料未提供具体基准结果。","implications":"对于使用长任务代理的开发者，后台守护进程、会话附着与恢复、工作进程崩溃恢复等机制，可能降低持续任务被中断后的操作成本。材料同时表示该项目开源，并称可配合现代开放与闭源前沿模型使用。","nextStep":"建议后续关注开源仓库的实际安装与使用反馈，以及不同模型接入后的任务完成质量、恢复可靠性和资源消耗。尤其应核验其自我修改机制在真实编码任务中是否带来可重复的收益，而非仅停留在架构描述。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-08-06T02:06:37.192Z","sourceHash":"99c8e0a3a46cb61a","review":{"approved":true,"groundedness":96,"clarity":91,"duplicationRisk":12,"blockingIssues":[],"notes":["“编码代理运行框架”与来源中的“coding harness”基本对应；如需更贴近原文，可统一表述为“编码代理框架（harness）”。","“可能降低持续任务被中断后的操作成本”属于合理推断，且已使用“可能”限定；来源直接支持的事实是后台守护进程、会话附着/分离及崩溃后的恢复机制。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["产品更新","Hacker News 热门（buzzing.cc 中文翻译）"],"translations":{"zh-CN":{"title":"Prime Agent：一个具有自我改进能力的RLM代理","summary":"Prime Agent 是一个自我改进的编码代理，围绕递归语言模型（RLM）和持续框架（Continual Harness）两大抽象构建，将上下文视为变量、子代理委派视为 REPL 内的函数调用，并允许代理对其提示词、技能、记忆和子代理进行 CRUD 操作。它完全开源，可通过 curl 命令安装，支持与前沿模型即时使用，并具备后台守护进程、会话恢复、分支分叉和异步内核压缩等特性。","category":"产品更新","source":"primeintellect.ai","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent：一个具有自我改进能力的RLM代理 - Aioga AI资讯","description":"Prime Agent 是一个自我改进的编码代理，围绕递归语言模型（RLM）和持续框架（Continual Harness）两大抽象构建，将上下文视为变量、子代理委派视为 REPL 内的函数调用，并允许代理对其提示词、技能、记忆和子代理进行 CRUD 操作。它完全开源，可通过 curl 命令安装，支持与前沿模型即时使用，并具备后台守护进程、会话恢复、分支分叉...","url":"https://www.aioga.com/news/cmsgulpsm0hwaro5qcjgjlvgr/"},"en":{"title":"Prime Agent: An RLM Agent with Self-Improvement Capability","summary":"Prime Agent is a self-improving coding agent built around two major abstractions: the Recursive Language Model (RLM) and the Continual Harness. It treats context as variables, sub-agent delegation as function calls within a REPL, and allows the agent to perform CRUD operations on its prompts, skills, memory, and sub-agents. It is fully open-source, installable via curl command, supported for instant use with cutting-edge models, and equipped with features such as background daemons, session recovery, branch forking, and asynchronous kernel compression.","category":"Products","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: An RLM Agent with Self-Improvement Capability - Aioga AI News","description":"Prime Agent is a self-improving coding agent built around two major abstractions: the Recursive Language Model (RLM) and the Continual Harness. It treats context as variables, sub-...","url":"https://www.aioga.com/en/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:05.075Z"},"ja":{"title":"プライムエージェント:自己改善能力を持つRLMエージェント","summary":"プライムエージェントは、再帰的言語モデル(RLM)と継続的ハーネスという2つの抽象要素を中心に構築された自己改善型コーディングエージェントです。コンテキストを変数として扱い、サブエージェントをREPL内の関数呼び出しとして委任し、エージェントがプロンプト、スキルメモリ、サブエージェントに対してCRUD操作を行うことを可能にします。 完全にオープンソースで、curlコマンドでインストール可能で、フロンティアモデルでの即時利用をサポートし、バックグラウンドデーモン、セッションリカバリー、ブランチフォーク、非同期カーネル圧縮などの機能も備えています。","category":"製品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"プライムエージェント:自己改善能力を持つRLMエージェント - Aioga AIニュース","description":"プライムエージェントは、再帰的言語モデル(RLM)と継続的ハーネスという2つの抽象要素を中心に構築された自己改善型コーディングエージェントです。コンテキストを変数として扱い、サブエージェントをREPL内の関数呼び出しとして委任し、エージェントがプロンプト、スキルメモリ、サブエージェントに対してCRUD操作を行うことを可能にします。 完全にオープンソースで、c...","url":"https://www.aioga.com/ja/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:06.378Z"},"ko":{"title":"프라임 에이전트: 자기 개선 기능을 가진 RLM 요원입니다","summary":"Prime Agent는 재귀 언어 모델(RLM)과 연속 하네스(Continual Harnesses) 두 가지 추상 요소를 중심으로 설계된 자기 개선형 코딩 에이전트입니다. 이 에이전트는 문맥을 변수로 취급하고, 하위 에이전트를 REPL 내 함수 호출로 위임하며, 에이전트가 프롬프트, 스킬 메모리, 하위 에이전트에 대해 CRUD 연산을 수행할 수 있도록 합니다. 완전 오픈 소스이며, curl 명령어로 설치할 수 있고, 프론티어 모델과의 즉시 사용을 지원하며, 백그라운드 데몬, 세션 복구, 분기 포크, 비동기 커널 압축 등의 기능을 지원합니다.","category":"제품 업데이트","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"프라임 에이전트: 자기 개선 기능을 가진 RLM 요원입니다 - Aioga AI 뉴스","description":"Prime Agent는 재귀 언어 모델(RLM)과 연속 하네스(Continual Harnesses) 두 가지 추상 요소를 중심으로 설계된 자기 개선형 코딩 에이전트입니다. 이 에이전트는 문맥을 변수로 취급하고, 하위 에이전트를 REPL 내 함수 호출로 위임하며, 에이전트가 프롬프트, 스킬 메모리, 하위 에이전트에 대해 C...","url":"https://www.aioga.com/ko/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:14.473Z"},"es":{"title":"Agente Principal: Un agente RLM con capacidades de auto-mejora","summary":"Prime Agent es un agente de codificación auto-mejorador construido en torno a dos elementos abstractos: Modelos de Lenguaje Recursivo (RLM) y Aprovechamientos Continuos. Trata el contexto como variables, delega subagentes como llamadas a funciones dentro de REPL y permite a los agentes realizar operaciones CRUD en sus prompts, memorias de habilidades y subagentes. Es completamente de código abierto, puede instalarse mediante el comando curl, soporta el uso inmediato con modelos frontera y características como demonios en segundo plano, recuperación de sesión, bifurcaciones de bifurcación y compresión asíncrona del kernel.","category":"Productos","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Agente Principal: Un agente RLM con capacidades de auto-mejora - Aioga Noticias de IA","description":"Prime Agent es un agente de codificación auto-mejorador construido en torno a dos elementos abstractos: Modelos de Lenguaje Recursivo (RLM) y Aprovechamientos Continuos. Trata el c...","url":"https://www.aioga.com/es/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:15.577Z"},"fr":{"title":"Agent principal : Un agent RLM doté de capacités d’amélioration personnelle","summary":"Prime Agent est un agent de codage auto-améliorant construit autour de deux éléments abstraits : les modèles de langage récursifs (RLM) et les harnais continus. Il traite le contexte comme des variables, délègue les sous-agents comme appels de fonction au sein du REPL, et permet aux agents d’effectuer des opérations CRUD sur ses invites, ses mémoires de compétences et ses sous-agents. Il est entièrement open source, peut être installé via la commande curl, supporte une utilisation immédiate avec des modèles frontières, et propose des fonctionnalités telles que les démons en arrière-plan, la récupération de session, les forks de branchement et la compression asynchrone du noyau.","category":"Produits","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Agent principal : Un agent RLM doté de capacités d’amélioration personnelle - Aioga Actualités IA","description":"Prime Agent est un agent de codage auto-améliorant construit autour de deux éléments abstraits : les modèles de langage récursifs (RLM) et les harnais continus. Il traite le contex...","url":"https://www.aioga.com/fr/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:24.847Z"},"de":{"title":"Prime Agent: Ein RLM-Agent mit Selbstverbesserungsfähigkeiten","summary":"Prime Agent ist ein selbstverbessernder Coding-Agent, der auf zwei abstrakten Elementen basiert: Rekursive Sprachmodelle (RLM) und kontinuierliche Harnesses. Er behandelt Kontext als Variablen, delegiert Unteragenten als Funktionsaufrufe innerhalb von REPL und ermöglicht es Agenten, CRUD-Operationen auf seine Prompts, Skill-Erinnerungen und Unteragenten auszuführen. Es ist vollständig Open Source, kann über den Befehl curl installiert werden, unterstützt die sofortige Nutzung mit Frontier-Modellen und Funktionen wie Hintergrund-Daemons, Session-Wiederherstellung, Branch-Forks und asynchrone Kernel-Kompression.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: Ein RLM-Agent mit Selbstverbesserungsfähigkeiten - Aioga KI-News","description":"Prime Agent ist ein selbstverbessernder Coding-Agent, der auf zwei abstrakten Elementen basiert: Rekursive Sprachmodelle (RLM) und kontinuierliche Harnesses. Er behandelt Kontext a...","url":"https://www.aioga.com/de/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:23.872Z"},"pt-BR":{"title":"Prime Agent: um agente RLM com capacidade de autoaperfeiçoamento","summary":"Prime Agent é um agente de codificação autoaperfeiçoável, construído em torno de duas abstrações principais, Modelo de Linguagem Recursivo (RLM) e Estrutura Contínua (Continual Harness). Ele trata o contexto como variáveis, as delegações de subagentes como chamadas de funções dentro de um REPL e permite que o agente realize operações CRUD em seus prompts, habilidades, memória e subagentes. É totalmente open source, pode ser instalado via comando curl, suporta uso imediato com modelos de ponta e conta com recursos como daemon em segundo plano, restauração de sessões, ramificação, bifurcação e compressão assíncrona do kernel.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: um agente RLM com capacidade de autoaperfeiçoamento - Aioga Notícias de IA","description":"Prime Agent é um agente de codificação autoaperfeiçoável, construído em torno de duas abstrações principais, Modelo de Linguagem Recursivo (RLM) e Estrutura Contínua (Continual Har...","url":"https://www.aioga.com/pt-BR/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:32.431Z"},"ru":{"title":"Prime Agent: агент RLM с возможностями самосовершенствования","summary":"Prime Agent — это саморазвивающийся кодировочный агент, построенный на двух абстрактных элементах: рекурсивных языковых моделей (RLM) и непрерывных угловых угловых элементах. Он рассматривает контекст как переменные, делегирует подагенты как вызовы функций внутри REPL и позволяет агентам выполнять операции с CRUD-обработкой своих запросов, памяти навыков и подагентов. Он полностью открыт по исходному коду, может быть установлен через команду curl, поддерживает немедленное использование с фронтирными моделями, а также такими функциями, как фоновые демоны, восстановление сессий, ветвления и асинхронное сжатие ядра.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: агент RLM с возможностями самосовершенствования - Aioga Новости ИИ","description":"Prime Agent — это саморазвивающийся кодировочный агент, построенный на двух абстрактных элементах: рекурсивных языковых моделей (RLM) и непрерывных угловых угловых элементах. Он ра...","url":"https://www.aioga.com/ru/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:34.534Z"},"ar":{"title":"الوكيل الرئيسي: وكيل RLM يتمتع بقدرات تحسين ذاتي","summary":"الوكيل الرئيسي هو وكيل ترميز ذاتي التحسين مبني حول عنصرين مجردين: نماذج اللغة التكرارية (RLM) والأدوات المستمرة. يعامل السياق كمتغيرات (REPL)، ويفوض الوكلاء الفرعيين كاستدعاءات دوال داخل REPL، ويسمح للوكلاء بتنفيذ عمليات CRUD على التوجيهات، وذاكرات المهارات، والوكلاء الفرعيين. هو مفتوح المصدر بالكامل، ويمكن تثبيته عبر أمر curl، ويدعم الاستخدام الفوري مع نماذج الحدود، وميزات مثل الشياطئ الخلفية، واستعادة الجلسة، وفروع الفروع، وضغط النواة غير المتزامن.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"الوكيل الرئيسي: وكيل RLM يتمتع بقدرات تحسين ذاتي - Aioga أخبار الذكاء الاصطناعي","description":"الوكيل الرئيسي هو وكيل ترميز ذاتي التحسين مبني حول عنصرين مجردين: نماذج اللغة التكرارية (RLM) والأدوات المستمرة. يعامل السياق كمتغيرات (REPL)، ويفوض الوكلاء الفرعيين كاستدعاءات دوا...","url":"https://www.aioga.com/ar/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:43.705Z"},"hi":{"title":"प्राइम एजेंट: आत्म-सुधार क्षमताओं वाला एक आरएलएम एजेंट","summary":"प्राइम एजेंट एक स्व-सुधार करने वाला कोडिंग एजेंट है जो दो अमूर्त तत्वों के आसपास बनाया गया है: रिकर्सिव लैंग्वेज मॉडल (आरएलएम) और कंटीन्यूअस हार्नेस। यह संदर्भ को चर के रूप में मानता है, उप-एजेंटों को आरईपीएल के भीतर फ़ंक्शन कॉल के रूप में सौंपता है, और एजेंटों को अपने संकेतों, कौशल यादों और उप-एजेंटों पर सीआरयूडी संचालन करने की अनुमति देता है। यह पूरी तरह से खुला स्रोत है, कर्ल कमांड के माध्यम से स्थापित किया जा सकता है, फ्रंटियर मॉडल के साथ तत्काल उपयोग का समर्थन करता है, और पृष्ठभूमि डेमॉन, सत्र पुनर्प्राप्ति, शाखा कांटे और अतुल्यकालिक कर्नेल संपीड़न जैसी सुविधाएं।","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"प्राइम एजेंट: आत्म-सुधार क्षमताओं वाला एक आरएलएम एजेंट - Aioga AI समाचार","description":"प्राइम एजेंट एक स्व-सुधार करने वाला कोडिंग एजेंट है जो दो अमूर्त तत्वों के आसपास बनाया गया है: रिकर्सिव लैंग्वेज मॉडल (आरएलएम) और कंटीन्यूअस हार्नेस। यह संदर्भ को चर के रूप में मान...","url":"https://www.aioga.com/hi/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:43.897Z"},"it":{"title":"Agente Principale: Un agente RLM con capacità di auto-miglioramento","summary":"Prime Agent è un agente di codifica auto-migliorante costruito attorno a due elementi astratti: Modelli di Linguaggio Ricorsivi (RLM) e Harnesses Continui. Tratta il contesto come variabili, delega i sotto-agenti come chiamate di funzione all'interno di REPL e consente agli agenti di eseguire operazioni CRUD sui suoi prompt, memorie delle competenze e sotto-agenti. È completamente open source, può essere installato tramite il comando curl, supporta l'uso immediato con modelli frontier e offre funzionalità come demoni in background, recupero sessione, branch fork e compressione kernel asincrona.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Agente Principale: Un agente RLM con capacità di auto-miglioramento - Aioga Notizie IA","description":"Prime Agent è un agente di codifica auto-migliorante costruito attorno a due elementi astratti: Modelli di Linguaggio Ricorsivi (RLM) e Harnesses Continui. Tratta il contesto come...","url":"https://www.aioga.com/it/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:52.567Z"},"nl":{"title":"Prime Agent: Een RLM-agent met zelfverbeteringsmogelijkheden","summary":"Prime Agent is een zelfverbeterende codeeragent die is opgebouwd rond twee abstracte elementen: Recursive Language Models (RLM) en Continual Harnesses. Het behandelt context als variabelen, delegeert subagenten als functieaanroepen binnen REPL, en stelt agenten in staat CRUD-operaties uit te voeren op zijn prompts, vaardigheidsgeheugens en subagenten. Het is volledig open source, kan worden geïnstalleerd via het curl-commando, ondersteunt direct gebruik met frontier-modellen en functies zoals achtergronddaemons, sessieherstel, branch forks en asynchrone kernelcompressie.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: Een RLM-agent met zelfverbeteringsmogelijkheden - Aioga AI-nieuws","description":"Prime Agent is een zelfverbeterende codeeragent die is opgebouwd rond twee abstracte elementen: Recursive Language Models (RLM) en Continual Harnesses. Het behandelt context als va...","url":"https://www.aioga.com/nl/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:04:53.098Z"},"tr":{"title":"Prime Agent: Kendini geliştirme yeteneklerine sahip bir RLM ajanı","summary":"Prime Agent, iki soyut unsur etrafında inşa edilmiş kendini geliştiren bir kodlama ajanıdır: Recursive Language Models (RLM) ve Sürekli Harnesses. Bağlamı değişkenler olarak ele alır, alt ajanları REPL içinde fonksiyon çağrıları olarak devreder ve ajanların istemleri, yetenek hafızaları ve alt ajanları üzerinde CRUD işlemlerini gerçekleştirmesine olanak tanır. Tamamen açık kaynaklıdır, curl komutuyla kurulabilir, frontier modelleriyle anında kullanımı destekler ve arka plan daemonları, oturum kurtarma, branch forklar ve asenkron çekirdek sıkıştırma gibi özellikler sunar.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: Kendini geliştirme yeteneklerine sahip bir RLM ajanı - Aioga AI Haberleri","description":"Prime Agent, iki soyut unsur etrafında inşa edilmiş kendini geliştiren bir kodlama ajanıdır: Recursive Language Models (RLM) ve Sürekli Harnesses. Bağlamı değişkenler olarak ele al...","url":"https://www.aioga.com/tr/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:05:02.322Z"},"vi":{"title":"Prime Agent: Một tác nhân RLM có khả năng tự cải thiện","summary":"Prime Agent là một tác nhân lập trình tự cải tiến được xây dựng dựa trên hai yếu tố trừu tượng: Mô hình Ngôn ngữ Đệ quy (RLM) và Dây điều khiển liên tục. Nó xử lý ngữ cảnh như các biến, ủy quyền các tác nhân con như các lệnh gọi hàm trong REPL, và cho phép tác nhân thực hiện các thao tác CRUD trên các prompt, bộ nhớ kỹ năng và các tác nhân con. Nó hoàn toàn mã nguồn mở, có thể cài đặt qua lệnh curl, hỗ trợ sử dụng ngay lập tức với các mô hình frontier, và các tính năng như daemon nền, phục hồi phiên, phân nhánh và nén kernel bất đồng bộ.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: Một tác nhân RLM có khả năng tự cải thiện - Tin tức AI Aioga","description":"Prime Agent là một tác nhân lập trình tự cải tiến được xây dựng dựa trên hai yếu tố trừu tượng: Mô hình Ngôn ngữ Đệ quy (RLM) và Dây điều khiển liên tục. Nó xử lý ngữ cảnh như các...","url":"https://www.aioga.com/vi/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:05:02.020Z"},"id":{"title":"Prime Agent: Sebuah agen RLM dengan kemampuan peningkatan diri","summary":"Prime Agent adalah agen pengkodean yang dapat meningkatkan diri, dibangun di sekitar dua abstraksi utama, yaitu Recursive Language Model (RLM) dan Continual Harness. Model ini melihat konteks sebagai variabel, delegasi sub-agen sebagai panggilan fungsi di dalam REPL, dan memungkinkan agen melakukan operasi CRUD pada prompt, keterampilan, memori, dan sub-agen. Agen ini sepenuhnya open source, dapat diinstal melalui perintah curl, mendukung penggunaan dengan model mutakhir secara langsung, dan memiliki fitur seperti daemon latar belakang, pemulihan sesi, percabangan, dan kompresi kernel asinkron.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: Sebuah agen RLM dengan kemampuan peningkatan diri - Berita AI Aioga","description":"Prime Agent adalah agen pengkodean yang dapat meningkatkan diri, dibangun di sekitar dua abstraksi utama, yaitu Recursive Language Model (RLM) dan Continual Harness. Model ini meli...","url":"https://www.aioga.com/id/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:05:10.213Z"},"th":{"title":"Prime Agent: ตัวแทน RLM ที่มีความสามารถในการพัฒนาตนเอง","summary":"Prime Agent เป็นเอเจนต์เขียนโค้ดที่พัฒนาตัวเอง สร้างขึ้นจากสององค์ประกอบนามธรรม ได้แก่ Recursive Language Models (RLM) และ Continual Harnesses มันจัดการบริบทเป็นตัวแปร มอบหมายซับเอเจนต์เป็นฟังก์ชันเรียกภายใน REPL และอนุญาตให้เอเจนต์ดําเนินการ CRUD กับพรอมต์ หน่วยความจําทักษะ และซับเอเจนต์ เป็นโอเพ่นซอร์สเต็มรูปแบบ สามารถติดตั้งผ่านคําสั่ง curl รองรับการใช้งานทันทีกับโมเดล Frontier และฟีเจอร์ต่าง ๆ เช่น background daemons, session recovery branch forks และการบีบอัดเคอร์เนลแบบอะซิงโครนัส","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: ตัวแทน RLM ที่มีความสามารถในการพัฒนาตนเอง - ข่าว AI Aioga","description":"Prime Agent เป็นเอเจนต์เขียนโค้ดที่พัฒนาตัวเอง สร้างขึ้นจากสององค์ประกอบนามธรรม ได้แก่ Recursive Language Models (RLM) และ Continual Harnesses มันจัดการบริบทเป็นตัวแปร มอบหมายซับเอ...","url":"https://www.aioga.com/th/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:05:11.685Z"},"pl":{"title":"Prime Agent: Agent RLM z możliwością samodoskonalenia","summary":"Prime Agent to samodoskonalący się agent programujący oparty na dwóch abstrakcyjnych elementach: Rekurencyjnym Modelach Językowych (RLM) oraz Continual Harnesses. Traktuje kontekst jako zmienne, deleguje subagentów jako wywołania funkcji w REPL i pozwala agentom wykonywać operacje CRUD na swoich promptach, pamięci umiejętności i subagentach. Jest w pełni open source, można go zainstalować za pomocą polecenia curl, wspiera natychmiastowe użycie z modelami frontier oraz funkcje takie jak daemony w tle, odzyskiwanie sesji, rozgałęzienia i asynchroniczna kompresja jądra.","category":"产品更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Prime Agent: Agent RLM z możliwością samodoskonalenia - Aioga Wiadomości AI","description":"Prime Agent to samodoskonalący się agent programujący oparty na dwóch abstrakcyjnych elementach: Rekurencyjnym Modelach Językowych (RLM) oraz Continual Harnesses. Traktuje kontekst...","url":"https://www.aioga.com/pl/news/cmsgulpsm0hwaro5qcjgjlvgr/","contentTranslated":true,"sourceHash":"7fbc07828c44b266","translatedAt":"2026-08-06T02:05:20.145Z"}}}}