我们正在构建一个原生人工智能质量保证平台——在竞争激烈的质量保证领域,你将如何发展?

1作者: pratik-tgx9 个月前原帖
我们是一个小型自筹资金的团队,已经在质量保证(QA)咨询领域深耕了一段时间。在多个项目中,我们发现了一些重复出现的问题: - 测试生成过程缓慢且手动 - 维护困难(尤其是易失的选择器) - 报告有时不尽如人意 - 易失的测试削弱了对自动化的信心 因此,我们开始构建一个工具,它能够: - 观察浏览器交互并生成 Node.js 测试代码 - 利用人工智能(包括视觉模型)实现自我修复和稳定选择器 - 目前正在增加大语言模型(LLM)和自然语言支持 我们与几位管理手动 QA 工程师团队的首席技术官(CTO)进行了交流。一个反复出现的主题是,他们希望赋能这些团队转向自动化,但被声称能做到这一点的工具市场的拥挤所压倒。 我们的挑战在于,尽管我们认为我们的工具确实可以提供帮助,但这个领域竞争异常激烈: 1) 无代码/低代码工具:Tricentis Tosca、Leapwork、Reflect、ACCELQ、Katalon,还有很多其他工具 2) AI 原生初创公司(许多得到 YC 支持):Browser Use、Momentic、Meticulous、Fix AI、Spur、Browser Base、Intryc、Skyvern、BugBug 等等 3) 机器人流程自动化(RPA):UIPath、Automation Anywhere、Blueprism 4) 还有大量提供捆绑服务和产品的 QA 咨询公司 即使有这么多工具,我们发现复杂的、特定产品的测试用例仍然会被忽视。这就是为什么许多团队仍然依赖 Selenium、Cypress 或 Playwright 来构建自定义框架——尤其是在需要更深层逻辑、API 集成、CI/CD 工作流或自定义部署处理时。 —— 我们目前面临的困境: 1) 我们生成实际代码,因此我们的理想用户是 SDET、QA 自动化工程师或开发人员 2) 但我们不想抛弃那些希望进行自动化的手动测试人员 3) 我们现在应该在此基础上构建一个无代码的用户界面层,还是先专注于技术用户? 此外,从市场推广的角度来看: 4) 我们应该开源基础工具以建立开发者信任吗? 5) 我们应该从一开始就选择闭源 SaaS 吗? 6) 还是将其产品化为由我们内部工具驱动的 QA 服务,比如 QA Wolf、Muuktest 或 Rainforest QA? 目前,我们正在积极采访 CTO、工程副总裁和高级工程师,以了解我们可以提供最大价值的地方——以及这是否是他们真正愿意购买的东西。 如果您曾在竞争激烈的 B2B 领域构建过开发工具、QA 平台或其他产品: - 您今天会如何处理这个问题? - 您会优先考虑开发人员还是手动测试人员? - 您会选择从开源、SaaS 还是产品化服务开始? 感谢您的阅读!
查看原文
We’re a small, bootstrapped team that’s been deep in QA consulting for a while. Across projects, we’ve seen the same issues repeat: • Test generation is slow and manual • Maintenance is painful (especially flaky selectors) • Reporting sometime sucks • Flaky tests kill confidence in automation<p>So we started building something that: • Observes browser interactions and generates Node.js test code • Uses AI (including vision models) for self-healing + stable selectors • Is now adding LLM + natural language support on top<p>We’ve spoken with several CTOs who manage teams of manual QA engineers. A recurring theme is that they want to empower those teams to move into automation, but are overwhelmed by the crowded landscape of tools claiming to do that.<p>Our challenge is that while we think we can genuinely help with our tool, the space is incredibly crowded: 1) No-code&#x2F;low-code tools: Tricentis Tosca, Leapwork, Reflect, ACCELQ, Katalon, so many others 2) AI-native startups (many YC-backed): Browser Use, Momentic, Meticulous, Fix AI, Spur, Browser Base, Intryc, Skyvern, BugBug, and ton of them. 3) RPA: UIPath, Automation Anywhere, Blueprism 4) Plus tons of QA consulting companies offering bundled services and their products<p>Even with all these tools, we’ve found that complex, product-specific test cases still fall through the cracks. That’s why many teams still rely on Selenium, Cypress, or Playwright to build custom frameworks — especially when they need deeper logic, API integration, CI&#x2F;CD workflows, or custom deployment handling.<p>⸻<p>Where we’re stuck: 1) We generate actual code, so our ideal user is SDETs, QA automation engineers, or developers 2) But we don’t want to leave behind manual testers who want to do automation 3) Should we build a no-code UI layer on top now — or focus on technical users first?<p>Also, from a go-to-market perspective: 4) Should we open source the base tool and build developer trust? 5) Should we go closed-source SaaS from the beginning? 6) Or productize as a QA service powered by our internal tool, like QA Wolf, Muuktest or Rainforest QA?<p>Right now, we’re actively interviewing CTOs, VPs of Engineering, and senior engineers to understand where we can deliver the most value — and whether this is something they’d actually buy.<p>If you’ve built a devtool, QA platform, or anything in a competitive B2B space: • How would you approach this today? • Would you prioritize devs or manual testers? • Would you start with open-source, SaaS, or productise service?<p>Thanks for reading!