NVIDIA AVO在ARC-AGI-3满分，TrueFoundry开源Age
🤖 AI Agents Weekly: NVIDIA AVO, TrueForge, Chroma Foundation, Fragile Self-Improvement, Ornith-1.5, dots3-note, DeepSeek Vision, and More
AVO证明了系统架构对长程推理能力的决定性作用，TrueForge则提供了可落地的低成本Agent运行时方案，两者均具高参考价值。
In today’s issue:
本期内容：
- NVIDIA AVO hits 100% on ARC-AGI-3
- TrueFoundry open-sources TrueForge harness
- Chroma ships Foundation agent memory
- Self-improvement gains vanish when reshuffled
- Ornith-1.5 self-improves to Opus level
- dots3-note runs for days
- DeepSeek adds vision to V4-Flash
- Stripe acquires OpenRouter
- Cursor rebuilds Git storage as a database
- Harvey post-trains Tenet on Kimi K3
- Slack Code makes coding multiplayer
- LEGO-RL trains agents in Claude Code
- NVIDIA AVO 在 ARC-AGI-3 上取得 100% 成绩
- TrueFoundry 开源 TrueForge 框架
- Chroma 推出基础代理记忆功能
- 重新洗牌后，自我提升的收益消失
- Ornith-1.5 自我提升至 Opus 水平
- dots3-note 持续运行数天
- DeepSeek 为 V4-Flash 增加视觉能力
- Stripe 收购 OpenRouter
- Cursor 将 Git 存储重构为数据库
- Harvey 在 Kimi K3 上对 Tenet 进行后训练
- Slack Code 让编程支持多人协作
- LEGO-RL 在 Claude Code 中训练代理
And all the top AI dev news, papers, and tools.
以及所有顶级 AI 开发新闻、论文和工具。
Top Stories
头条新闻
NVIDIA AVO Solves ARC-AGI-3
NVIDIA AVO 解决 ARC-AGI-3
NVIDIA’s general-purpose coding agent AVO scored a perfect 100.00 RHAE on the ARC-AGI-3 public set, clearing all 183 levels across all 25 environments with no instructions, rules, or stated goals.
NVIDIA 的通用编码代理 AVO 在 ARC-AGI-3 公开集上取得了完美的 100.00 RHAE 分数，无需任何指令、规则或明确目标，便成功通过了所有 25 个环境中的全部 183 个关卡。
- The harness carries the result: Claude Opus 5 alone scores roughly 30% on the same benchmark. Wrapping it in AVO takes it to 100%, which NVIDIA frames as evidence that system design, not model capability alone, unlocks frontier long-horizon performance.
- Agentic variation loop: AVO cycles through inspect context, plan the next change, implement, evaluate with a scoring function, then diagnose and repair from failed attempts, committing accepted candidates into a growing solution lineage.
- Persistent memory and a supervisor: Prior implementations, evaluation results, compiler output, and reasoning are retained so the agent resumes from current state, while a separate monitor watches for stagnation and conditionally redirects the main agent.
- Transfer across domains: The same architecture was originally built for CUDA GPU kernel optimization and moved to interactive reasoning unchanged, using about 12% fewer environment actions than the prior VISTA baseline.
- 该框架得出的结果是：仅凭 Claude Opus 5 在同一基准测试中得分约为 30%。将其包裹在 AVO 中后，得分提升至 100%，NVIDIA 将此视为证据，表明系统架构设计而非单纯的模型能力，才是解锁前沿长程性能的关键。
- 代理式变分循环：AVO 循环执行检查上下文、规划下一步变更、实施、使用评分函数进行评估，然后从失败尝试中进行诊断和修复，并将被接受的候选方案提交到不断增长的解决方案谱系中。
- 持久化记忆与监督机制：保留先前的实现、评估结果、编译器输出和推理过程，以便代理从当前状态恢复；同时由独立的监控器观察停滞情况，并条件性地重定向主代理。
- 跨领域迁移：该架构最初是为 CUDA GPU 内核优化而构建的，随后未经修改地迁移至交互式推理任务，相比之前的 VISTA 基线，环境操作减少了约 12%。
Blog | Agent Security
博客 | 代理安全
TrueFoundry Open-Sources TrueForge
TrueFoundry 开源 TrueForge
TrueFoundry open-sourced TrueForge under an MIT license, a vendor-neutral agent harness that acts as the runtime layer turning an LLM into a working agent.
TrueFoundry 以 MIT 许可证开源了 TrueForge，这是一个厂商中立的代理框架，作为运行时层将大语言模型转化为可用的代理。
- Batteries included: MCP tools, a skills registry, sandboxed execution, human-in-the-loop approval gates for sensitive actions, subagents, durable state for long-running tasks, and step-level tracing all ship in the box.
- Vendor-neutral by design: It runs on your own infrastructure against any commercial or open-source model, with per-task switching, and the TrueFoundry AI Gateway is optional rather than required.
- Cost is the pitch: On DevRev’s Enterprise-Bench, evaluated blind, TrueFoundry reports roughly 30% lower cost on identical tasks with the same model, and up to 75% savings when routing to open-source models with accuracy matched.
- Traction: The repo has cleared 2,800 stars since the August 19 launch. Star it if you want to follow the project.
- 内置功能：MCP 工具、技能注册表、沙盒执行、敏感操作的人工审批关卡、子代理、长运行任务的持久状态以及步骤级追踪，均已开箱即用。
- 设计中立：它运行在您自己的基础设施上，支持对接任何商业或开源模型，具备按任务切换的能力，且 TrueFoundry AI Gateway 为可选而非必需。
- 成本优势是核心卖点：在 DevRev 的企业基准测试中（盲测），TrueFoundry 报告称在使用相同模型处理相同任务时，成本降低约 30%；而在路由至开源模型且保持准确率相当的情况下，节省可达 75%。
- 市场反响：自 8 月 19 日发布以来，该仓库已获得 2,800 个星标。如果您想关注该项目，请为其点亮星标。
GitHub | Docs | Product
GitHub | 文档 | 产品
Read more
阅读更多
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力