Automattic debuts a platform to deploy vibe coding apps amid a changing internet
WordPress has been a crucial technology and content management system for websites like TechCrunch. Now, WordPress.com owner Automattic is responding to the changing state of the web by debuting a new service called Spacefast, which will allow users to deploy vibe-coded websites and apps.
Spacefast is designed to provide infrastructure for the vibe-coded apps, including databases, authentication, file uploads, emails, and analytics, to make them functional websites, apps, or services. The product competes with the likes of Lovable and Emergent, which offer both interface and rails for vibe-coded apps, but in a more open way.
Users can connect Spacefast with other agents and chatbots like ChatGPT, Claude, Hermes, OpenClaw, Grokbot, Muse, Pi, Cursor, Opencode, and Amp. Alternatively, they can connect a Git repository to Spacefast to get started.
The product is headed by Batuhan İçöz, who worked on working on it during one of the company’s hackathon sprints earlier in the year. At that time, he worked on the company’s cross-platform messaging app Beeper and said that he often got Vercel links from his non-technical friends while exploring vibe coding.
“I thought there had to be a better way for everyone to set up and deploy their projects. Automattic already has good infrastructure, but we have been traditionally associated with just WordPress. I thought, why do I have to use Cloudflare when we have a CDN (Content Delivery Network) just as good? So I started working on what was called Stattic and now Spacefast,” he told TechCrunch over a call.
İçöz eventually disseminated the solution to his friends and added more features to make the product more dynamic. Automattic soft-launched the product in September to get feedback before today’s launch.
The company has seen varied use cases of their product, with people and organizations using it for internal mockups, slideshows, and offer-related websites. However, that’s not the only competition Spacefast has. Productivity and design software like Figma, Canva, and Notion have launched products that sit at the intersection of vibe coding, mockups, and presentations.
İçöz thinks that, compared to something like Lovable, Spacefast has the advantage of giving users a choice of their agent and using the solution their way. The counter-argument might be that a certain kind of user wouldn’t want to choose and just have everything figured out for them.
Spacefast said that it also has the advantage of being within Automattic to launch solutions like web analytics, WordPress integration, and e-commerce features (through WooCommerce) easily to compete better with the whole lot.
“For enterprises, the problem right now is IT sprawl, as their marketing teams are churning out landing pages on services like Lovable, and they want to consolidate and work with a company that understands the enterprise needs. With our existing products, we have experience serving enterprise customers,” İçöz noted.
A challenge for newly minted products in the AI world is to retain people post-launch. Users might get excited to try out a new product, and then move on to the new shiny thing within days or weeks. İçöz is cognizant of the fact and said that he expects people to churn, as they might not be interested in making apps at all.
“I think what we need to do as a product is make sure that when you build something with SpaceFast, it is so easy, and the results are so much better than without Spacefast,” İçöz emphesized.
Many websites and publishers are struggling with dwindling traffic due to AI chatbots and summaries, right now, to the point that Google itself saw its ad revenue growth drop. İçöz acknowledged that people are making websites with ChatGPT or Claude now as compared to deploying something on WordPress, and he said some sites won’t need a content management system at all as the web expands.
Spacefast comes amid turbulent months at Automattic. In an OpenAI-Sam Altman-style saga, the company’s board forced CEO Matt Mullenweg out, but he was back in control within days. The rebel board was eventually replaced, and the interim CFO left within weeks of appointment. In October, TechCrunch reported that the company had began layoffs across several properties and more cuts were expected.
https://techcrunch.com/2026/11/09/automattic-debuts-a-platform-to-deploy-vibe-coding-apps-amid-a-changing-internet/Reposted. Will be removed on request. All rights belong to the original author.
More from this batch
Our newest TPUs: 8tx and 8i Max, built for the agentic era
Google adds two eighth-generation TPU systems: TPU 8tx for hardware-level confidential inference and private fine-tuning, and the dual-chip …
Anthropic 最早周二启动 IPO 路演,上周 AI 股暴跌后估值讨论已降到约 1.1–1.3 万亿美元
Bloomberg Law 11 月 9 日报道:Anthropic 的等待期到期,最早周二就能正式向投资者推介 IPO;上周 AI 相关股票大跌,一度把波动率指数推到 25 以上,这个区间通常会让公司推迟上市。Bloomberg 11 月 6 日报道称,潜在买家在讨论 1.2–…
Tencent's AI chief says AI bioweapons are coming, but closing off models isn't the answer
Tencent chief AI scientist Yao Shunyu argues for a mix of open and closed models as AI nears bioweapon capability; The Decoder notes open-mo…
特朗普在《60 分钟》谈 AI 安全:风险「没有他们说的那么真」,还称司法部警告过一位大 AI 公司负责人「准备坐牢」
特朗普没有点名是哪家公司,只说是「大公司之一」,并称对方之后「彻底改变了一切」。他表示愿意和民主党合作搞 AI 安全护栏,但不想让美国落到「第二」。在民主党刚赢得国会、AI 监管预期升温的背景下,这段表态值得留意。
迪士尼和 Adobe 签多年合作:用 Firefly Foundry 训练专属迪士尼角色和画风的生成模型
Adobe 在 MAX 大会宣布,双方会在 Firefly Foundry 上做「以艺术家为主导」的定制模型,用迪士尼自己的角色、场景和视觉风格训练,输出可商用;迪士尼的主题乐园、邮轮和 ESPN 都会提前用上 Adobe 新技术。The Verge 指出,此前彭博报道迪士尼正准…
CoreWeave 和 Wayve 签多年协议:Wayve 的自动驾驶基础模型在 CoreWeave 云上训练,新模型不比现役强就不上路
CoreWeave 官方宣布,Wayve 会在它的平台上追踪、评测和对比每一个 AI Driver 模型,只有超过当前上路版本的才会部署。这说明「物理 AI」正在成为 GPU 云的新需求来源,不只是大语言模型。
美团 LongCat-2.5 正式开放权重,MIT 协议可商用:1.6 万亿参数、每 token 激活约 480 亿、百万上下文
这是 11 月 2 日发布的模型这次真正放出权重,Hugging Face 上同时有 W4A4 版本。改进点:视觉信号被编码成多层级离散 token,和文本走同一个主干,而不是外挂视觉模块,所以视觉编程和 3D 类任务更强;推理时专家权重和激活都量化到 4 bit,直接用 INT…
Amazon 提出 HOPE 专家剪枝:考虑专家之间的「搭档关系」,砍掉一半专家时智能体编程任务比次优方法最多高 6.1%
问题:MoE 模型推理时所有专家都要占显存,比如 Qwen3.5-122B 约 95% 的参数内存都在专家 MLP 上,但每个 token 只用 8 个。现有剪枝给每个专家单独打分,忽略了专家会成组协作。HOPE 用二阶方法考虑专家两两之间的交互,在 50% 剪枝率下平均排名最好…
LowRAM:同一注意力头里的 V 大多能从 K 预测出来,据此把 KV 缓存压到 4 倍近无损,叠加量化后可压 26.9 倍
改进了什么:长上下文时 KV 缓存的显存随长度线性增长,限制了批量和吞吐。靠什么:作者发现每个头的 value 大体可以由 key 预测,于是测试时拟合一个轻量预测器,只给重建最差的 token 存残差;再配上跨层低秩 key 压缩。效果:4 倍压缩下解码峰值吞吐最高提升 5 倍…
英伟达 AutoCompress:按目标显卡和推理引擎的真实延迟、吞吐、显存来剪枝,把 Nemotron 3.5 Super 从 120B 压到 74B 单卡 B200 跑 128K
改进了什么:以前剪枝按参数量挑大小,但参数量预测不了实际延迟和显存。靠什么:把推理引擎的成本模型放进可微架构搜索,同时优化全局和逐层结构,得到每层不一样的异构架构。效果:比 Minitron 在 RTX 5070、5080 和 H100 上平均准确率分别高 6.57、2.35 和…
FinDendro:大模型「不知道自己知道什么」,用它做金融回测没法靠自报或知识截止日期保证没有信息泄露
新基准包含 1999–2026 年 2,226 条带日期的美国宏观、股价和美联储决议事实。结果:多数前沿模型低估自己的知识,GPT-5 答对 76%,却只声称知道 38%,在它说「不知道」的事实里也答对 61%;让模型「假装站在过去某天」几乎不改变准确率。结论是,在训练窗口内的回…
Nace AI 的 Drex Max v1.1 决策模型登上 Artificial Analysis 决策榜第一,延迟 183 毫秒,即将开放权重
Drex Max 不生成文本,而是给定一组选项和上下文直接做选择,支持 256K 上下文。kimmonismus 转述论文:延迟约 183 毫秒,对比 Claude Opus 5.5 约 3.5 秒、GPT-6 Sol 约 2 秒。对路由、分类、审批这类企业决策环节,这种「只做选…