Factory 2.0: From coding agents to software factories
Factory announces its 2.0 release, repositioning from individual AI coding agents to an end-to-end 'software factory'. The post argues that improving individual productivity is no longer enough; enterprises need an interconnected, agent-native system that forms a continuous feedback loop from signals (bug reports, customer feedback) through planning, building, testing, reviewing, securing, shipping, and monitoring. Key design principles include model independence (allowing deliberate model choice or automatic routing per task), sovereign intelligence (data plane and control plane options from cloud to fully air-gapped, with all agent sessions and reviews feeding back into the system), and continual learning and self-improvement across the lifecycle. The article lists customers such as NVIDIA, EY, Adobe, and Palo Alto Networks already running software factories in production. Autonomy is described as a gradual maturation process, using simple Droids, skills, automations, Droid Computers, and multi-agent Missions for different levels of human guidance and agent readiness. The piece is a product launch announcement with some technical concepts, targeting engineers and managers interested in enterprise AI engineering and agent orchestration.
In 2023, we launched Factory with the mission to bring autonomy to software engineering. While others were using models to speed up coding, we set out to deploy autonomous Droids across the enterprise software development lifecycle. Today we are announcing the next phase in the future of software engineering.
Improving the productivity of individual engineers is no longer enough. Unlocking organization-wide productivity requires an interconnected, agent-native, end-to-end system. This system must improve over time by observing itself. The incremental units of this system are AI agents. This system will be built by engineers, and in turn will build their software.
This system is the software factory.
2023 年，我们创立了 Factory，使命是为软件工程带来自主性。当其他公司正用模型加速编码时，我们开始在企业的软件开发生命周期中部署自主 Droid。今天我们要宣布软件工程的下一个阶段。
仅提升单个工程师的生产力已远远不够。要实现组织级的生产力跃升，需要一个互联互通、以 Agent 为中心、端到端的系统。这个系统必须通过自我观察来持续改进。其增量单元是 AI 代理。这套系统将由工程师构建，反过来，它也将构建工程师的软件。
这个系统就是软件工厂。
The software factory starts with signals from the outside world: bug reports, internal conversations, customer feedback, business requirements. These signals get triaged and turned into planned changes. These changes are built, tested, reviewed, secured, shipped, and monitored. Monitoring that deployed software generates more signals. The entire system is a continuous feedback loop. Almost no one has meaningfully instrumented this loop to be fully AI-driven. We are still early, but the proliferation of software factories is going to happen very quickly.
软件工厂的起点来自外部世界的信号：Bug 报告、内部对话、客户反馈、业务需求。这些信号经分诊处理，转化为计划中的变更。这些变更经过构建、测试、审查、安全加固、发布和监控。对已部署软件的监控又会生成更多信号。整个系统就是一个持续反馈的闭环。目前几乎还没有人真正将这个回路完全用 AI 驱动起来。我们仍处于早期，但软件工厂的普及将会非常迅速。
A robust software factory must have Model Independence. Every model has a different trade-off of cost, performance, and speed. No one model fits every need within an enterprise. Your software factory must allow your organization to deliberately choose different models, or rely on a Router to automatically (or rule-based) select the best* model for any given task. As models commoditize, costs decrease while speed and performance increase.
一个健壮的软件工厂必须具备模型无关性。每个模型在成本、性能和速度上都有不同的取舍，没有一种模型能适合企业内部的所有需求。你的软件工厂必须允许组织有意识地选择不同模型，或者依赖一个 Router 来自动（或基于规则）为任意任务选择最佳*模型。随着模型商品化，成本在降低，速度和性能则在提升。
Sovereign Intelligence. You must be the sovereign of your software factory. Whether fully hosted in the cloud, bring-your-own-key, self-hosted data plane, EU-specific, or completely air-gapped with no external network access. Sovereignty means more than choosing where the system runs. It means owning a system that learns from itself, feeding every agent session, code review, and resolved incident back into the loop. The more you invest in your software factory, the more capable it becomes, and that capability stays with you, inside your walls, under your control.
主权智能。你必须是软件工厂的主权者。无论是完全托管在云端、自带密钥、自托管数据平面、针对欧盟部署，还是在完全离线的气隙环境中运行，主权意味着你不仅可以选择系统的运行位置，更意味着你拥有一个能从自身学习的系统——每一次 Agent 会话、代码评审、已解决的故障都会被反馈到回路中。你对软件工厂投入越多，它的能力就越强，而这份能力始终留在你手中、留在你的围墙内、始终受你掌控。
Continual Learning and Self-Improvement. Every stage of the software development lifecycle must be instrumented. When code review, security analysis, documentation, QA, and incident response all run on the same platform, they share the same agent core, the same model router, the same organizational context. A security finding informs the code review. A deployment triggers a documentation update. An incident correlates with the PR that caused it. Every additional automation, integration, or customization flows to the entire organization at once. The router itself learns to optimizes resources. The assembly line should span the full floor of the software factory.
持续学习与自我改进。软件开发生命周期的每个阶段都必须被仪表化。当代码审查、安全分析、文档、质量保证和故障响应都在同一个平台上运行时，它们共享同一个 Agent 核心、同一个模型路由、同一个组织上下文。一个安全发现会反馈到代码审查中；一次部署会触发文档更新；一个故障会与导致它的 PR 相关联。每当增加一项自动化、集成或定制化能力，整个组织都能立刻受益。路由本身也会学习如何优化资源。装配线应覆盖软件工厂的全流程。
We have been building software factories with our customers for the last few months. Software factories are already in production across the world’s largest organizations including NVIDIA, EY, Adobe, Palo Alto Networks, Adyen, Blackstone, Wipro, Comarch and more.
过去几个月，我们一直与客户共同打造软件工厂。目前软件工厂已在全球多家大型企业投入生产，包括 NVIDIA、EY、Adobe、Palo Alto Networks、Adyen、Blackstone、Wipro、Comarch 等。
No organization starts with a fully autonomous software factory. Autonomy is a maturation process that is gradual and specific to every organization’s readiness and comfort level. Deploying autonomy across the organization happens through deliberate engineering effort to codify workflows and standardize processes.
Factory enables a spectrum of autonomy over time. Not every process should use long-horizon autonomous tasks. Well-defined, measurable tasks run with simple Droid agents or skills. Automations coordinate recurring workflows with a shared objective and memory. Remote and persistent execution leverages Droid Computers for long running or local agents. Multi-agent autonomous execution called Missions solve complex tasks over hours or days by decomposing work into parallel tracks to handle. Different autonomous processes are used for varying requirements based on the level of human guidance required, the information sensitivity, and the level of Agent Readiness.
没有任何组织一开始就拥有完全自主的软件工厂。自主性是一个逐步成熟的过程，具体取决于每个组织的准备程度和舒适度。在组织内部署自主性需要经过精心的工程努力，将工作流规范化、标准化。
Factory 随时间推移提供一系列自主程度。并非所有流程都需要长周期自主任务。定义清晰、可衡量的任务由简单的 Droid 代理或技能来处理。自动化通过共享目标和记忆来协调重复性工作流。远程和持久化执行利用 Droid Computers 来运行长时间运行或本地的代理。多 Agent 自主执行称为 Missions，通过将复杂任务分解为并行的轨道来处理，耗时数小时或数天。根据所需的人为指导程度、信息敏感度以及 Agent 就绪度，会采用不同的自主流程来满足不同的需求。
Organizations that invest in their autonomous software development will see engineering outcomes surge, while owning decisions around cost, quality and context. The role of engineers is all the more important in this new era. No longer will they be the sole custodians of building the software. Instead, they will be responsible for building the factories that build the software. With this comes the responsibility of governance, safety, and the ownership of business outcomes. The next era of software development will be engineering-led and will see engineering responsibilities grow to span across the business itself.
Software factories are not built in a day, but the best day to start building your Factory is today.
Today we are expanding this functionality with visibility to manage your software factory directly in the Factory Desktop App.
那些在自主软件开发上投入的组织，将看到工程成果的爆发，同时也能掌控成本、质量和上下文。在这个新时代，工程师的角色反而更加重要。他们不再仅仅是软件构建的唯一守护者，而是将负责构建那些构建软件的工厂。随之而来的是治理、安全以及对业务成果的归属感。软件开发的下一阶段将由工程主导，工程师的职责将扩展至覆盖整个业务本身。
软件工厂不是一天建成的，但开启建设的最佳时机就是今天。
今天我们扩展了这一功能，通过 Factory Desktop App 可以直接管理你的软件工厂。
- best can be set according to cost, performance, speed, or some combination.
- 最佳可依据成本、性能、速度或某种组合来设定。