A seed, planted deliberately.
一粒种子,认真种下。
MTL Studio is a small independent studio building thoughtful AI software, end-to-end. It is also our experimental ground: every project exists to explore an idea and serve a real purpose, not to chase revenue.
麦团露工作室(MTL Studio)是家小型独立工作室,用心做 AI 软件,从想法到上线都自己动手。这里也是我们的试验田:每个项目都是为了试一个想法,做出来也得真有用,不是冲着赚钱去的。
Ink of Liangshan 墨染梁山
墨染梁山 Ink of Liangshan
It started somewhere without a cell signal. Every offline game we tried was either too heavy, draining the battery in an hour, or built around a story and a style we did not care for. So we made our own: a single-player tactics RPG retelling the Water Margin (水浒传) on a bright, hand-painted isometric board where every blow lands in ink, fifteen chapters and seventeen heroes, fully offline, borrowing the mechanics we loved from the tactics games we grew up on. The campaign is complete and the game is in TestFlight. This autumn it was repainted from the ground up, rebalanced and given an original soundtrack: four difficulties, a trait for every hero, five rarities of arms with a legendary weapon to hunt for each, and a promotion worth earning. It is now being playtested and hardened for its App Store release. It is also our experiment in AI-assisted game development: every system, painting and line of dialogue built end-to-end with AI in the loop. mtlstudio.net/mo-ran-liangshan
事情要从一个没信号的地方说起。当时试了一圈离线游戏,要么太耗电,一个小时就没电了,要么故事和画风都不对胃口。干脆自己做一款:单机战棋,重讲《水浒传》。战场是手绘的等距视角,色调明快,一招一式落下去都是一笔墨。十五回战役,十七位好汉,全程离线,玩法上借鉴了不少我们从小玩到大的老战棋。战役已经全部做完,目前在 TestFlight 测试。今年秋天整个游戏重画了一遍,数值重新调过,还配上了原创音乐:四档难度,好汉一人一绝,兵甲分五品,每位好汉还有一件专属神兵可寻,转职也得凭本事挣。眼下正在试玩、打磨,准备上架 App Store。这也是我们拿 AI 辅助做游戏的一次实验:系统、画面、对白,从头到尾都有 AI 参与。mtlstudio.net/mo-ran-liangshan
Reckoned Yet? 算了么?
算了么?Reckoned Yet?
A contemplative AI fortune-telling app rooted in BaZi (八字), the Chinese Four Pillars tradition. Bilingual from day one, built for English-speaking users and the Chinese-speaking diaspora. It is our experiment in attracting users and maximizing engagement. Version 1.0 launched on the web in October 2026 and runs in any browser. Readings answer first and stay short, sign-in works with a code sent by email, and profiles follow an account from one device to the next. app.reckoned.app
一款 AI 八字应用,气质偏沉静。从第一天起就是中英双语,做给英语用户和海外华人用。我们拿它来试怎么拉新、怎么把人留住。2026 年 10 月,1.0 版在网页端上线,打开浏览器就能用。解读先说结论,不绕弯子,篇幅也短;用邮箱验证码登录;档案跟着账号走,换台设备照样在。app.reckoned.app
AI Trading
AI 交易 AI Trading
A research project on language-model-driven trading systems and quantitative pipelines. It is our experiment in verifying market hypotheses and testing trading strategies, with an honest, evals-first approach to letting models touch markets. After a season of options-income and event-driven strategies in paper trading, it has become a moat-quality stock picker: a weekly report that ranks a universe of durable-advantage companies through sector-aware quality gates on their SEC filings, for holding as shares or long-dated calls. A paper account runs the same basket as its live track record, and a language model advises on market regime while the rules keep the final say.
一个研究项目,琢磨怎么用语言模型做交易系统和量化流程。我们拿它来验证对市场的假设、检验交易策略;让模型碰市场这件事,我们的态度是实事求是,先评测,再下场。在模拟盘上跑了一季期权收租和事件驱动策略之后,它现在成了选股工具,专挑护城河深的公司:每周出一份报告,根据各家公司提交给 SEC 的财报,分行业设质量门槛,给一批有长期优势的公司过筛、排序,拿来持股或者买长期看涨期权都行。另开了一个模拟账户,按同一篮子股票持仓,用来持续记录业绩;语言模型负责判断市场环境、给建议,最后拍板的还是规则。
We are a small studio building software for the long now. Every product begins as a question we want answered, and we build it end-to-end until the answer is real. We take on builds where the difference between almost working and actually working matters: senior engineers on the keyboard, from research through to your customers’ hands, and through the long tail of operations after launch.
我们是家小工作室,做软件图的是长久。每款产品起头都是一个我们自己想弄明白的问题,然后从头做到尾,做到答案立得住为止。有些项目,差不多能用和真的能用是两码事,我们接的就是这种:资深工程师亲自写代码,从前期研究一路做到产品交到你的客户手里,上线之后日复一日的运维也接着管。
Alongside our own products, we are open to a small amount of AI consulting and advisory work. Availability is limited, so we engage selectively, where our end-to-end experience is a genuine fit. If you are interested, write to contact@mtlstudio.net.
除了自家产品,我们也接少量 AI 咨询和顾问工作。人手有限,所以挑着接:我们这套从头做到尾的经验确实用得上,才接。有兴趣的话,欢迎来信contact@mtlstudio.net。
Curious about what we’re building? Say hello.
好奇我们在捣鼓什么?来打个招呼。
contact@mtlstudio.net