文 | 字母AI
Anthropic昨天公布了一笔特殊的交易：今年7月，它收购了一家名为Mithril Medicine的生物医药初创公司。这家公司只有一名全职员工，同时推进着两种罕见遗传病的药物研发项目，其中一种疗法已经在小鼠实验中取得初步成果。
公司的创始人利用Claude协调11家外部实验室，处理实验数据、项目安排和日常运营，独自维持着公司的运转。
从提出治疗构想到完成初步动物实验验证，只用了不到18个月。
收购消息公开当天，Mithril创始人Cody Tranbarger在X上发布了一篇长文：
“惊喜一：Anthropic收购了一家只有一个人的生物技术公司。”
“惊喜二：那个人就是我。”
“惊喜三：接下来我就来讲讲这件事……”
只有一个人的制药公司，被Anthropic看中
Anthropic没有公布这笔收购的金额，但明确表示，Mithril Medicine的两条药物研发管线，已经成为公司首批内部药物研发项目。
也就是说，Anthropic正在亲自参与新药研发，而这次被它收购的公司，只有一个全职员工。
当然，一家制药公司不可能只靠一个人完成所有实验。Mithril的背后还有生命科学风投机构Atlas Venture的资金支持、兼职科研顾问，以及11家负责实际实验的外部实验室。
创始人Cody负责确定研究方向、制定计划、判断实验结果、协调不同机构之间的合作，至于大量繁琐的日常工作，他交给了Claude——这可能也是Anthropic看中这家公司的原因之一。
根据Anthropic公布的研究资料，Claude承担了大量实验室协调、项目交接、数据整理和实验规划工作，甚至包括整理合同、追踪项目进度和起草工作邮件。
这些工作过去通常需要专门的运营和科研人员负责，如今Cody借助Claude，就能管理分散在不同机构的研发任务。Mithril因此得以长期维持只有一名全职员工的规模。
公司研发的两种疗法，分别针对LAMA2相关先天性肌营养不良症，以及α-肌营养不良蛋白病。这两种罕见遗传病都会破坏肌肉细胞与周围组织之间的连接，造成严重的肌肉无力和损伤。Mithril尝试用一种双特异性抗体，同时结合肌肉细胞和周围的支撑结构，像桥梁一样重新建立连接。
其中，针对LAMA2疾病的研究已经完成初步动物实验。接受最高剂量治疗的小鼠，握力和部分肌肉力量指标约为未治疗组的两倍。在实验观察期间，8只未治疗小鼠中有4只死亡，而最高剂量治疗组只有1只死亡。
对于Anthropic而言，Mithril的吸引力或许就在于此。它已经在真实的药物研发项目中尝试了一种特殊的工作模式：一名经验丰富的创业者，借助Claude协调外部科研资源，完成过去需要更多内部员工承担的工作。
Cody将这种模式称为“Pipeline-in-a-Person”（一人研发管线），Mithril被收购后，他继续负责原有的两条研发管线，并准备启动更多罕见病研究项目。
按照目前的计划，进展最快的项目有望在2028年底进入人体临床试验。
Gotta Cure 'Em All.
过去六年，Cody一直在X上使用“Professor Oak”（《宝可梦》里的大木博士）这个化名，分享生物医药行业的研究动态、投资观点和各种见闻。虽然这些年匿名程度越来越低，但他始终没有正式告诉大家自己是谁。
直到被Anthropic收购，他才决定主动揭晓身份。
他在文中自嘲，如今X上实在太无聊了，自己只好亲自曝光身份，给时间线增加一点乐趣。
Cody本科毕业于美国阿默斯特学院，主修生物学和经济学，大学期间还是学校长曲棍球队的守门员。毕业后，他先是进入高盛，从事生物医药企业的融资与并购业务，随后转向生命科学投资，在Casdin Capital工作。
2019年，Cody在了解哈佛商学院MBA项目时，偶然读到了校友Rich Horgan的个人介绍。Rich创办了一家名为Cure Rare Disease的罕见病公益组织，致力于为罕见遗传病患者开发个性化疗法，但其中一些疾病的患者人数太少，很难吸引传统药企投入研发。
接下来几个月，Cody一直惦记着这件事，最后干脆通过LinkedIn给Rich发了一条私信，表示自己愿意提供帮助。
两人就这样建立了联系。Cody逐渐参与组织的顾问工作，还帮忙联系企业赞助商。2020年，他参加了Cure Rare Disease举办的慈善晚会，后来又加入了组织的董事会。
在该组织的采访里，Cody提到，自己一直关注一种名为“n-of-1 medicine”的医疗模式，也就是为极少数患者、甚至单个患者开发个性化治疗方案。他希望让这些患者也能用上先进的治疗技术，并尝试降低此类疗法的研发成本。
此后，Cody还加入了基因编辑公司Prime Medicine，积累了更多生物技术企业的运营经验。
投资和创业的经历并非总是一帆风顺，Prime Medicine于2022年上市，后来股价持续走低。Cody在2025年发表的一篇文章中回忆，自己当年获得的员工股票期权后来全部深度价外，他最终没能赚到钱。但他依然很感激这段经历，许多前同事至今还是他的导师和朋友。
2023年，Cody进入哈佛大学，攻读生物技术MS/MBA双学位，并参与创办了生物技术公司Revision Bio。
到2025年，他又在Atlas Venture团队的帮助下创办了Mithril Medicine。
在10月27日的身份揭晓帖里，Cody专门感谢了Atlas Venture的Mike Gladstone、Jason Rhodes等人。他开玩笑说，自己在创业过程中做对了足够多的决定，终于让他们相信，Mithril和自己值得获得一笔种子投资。
“毕竟就连坏掉的钟一天都能准两次。”他说，自己能够获得这样的支持，实在值得感激。
在长文里，他还感谢了自己混迹多年的BioTwitter社区。Cody说，自己最初在X上潜水了六年，才开始以“大木博士”的身份发帖。过去十多年，这个平台带给他的东西相当多：行业知识、一大群朋友、许多工作机会、几位导师、一些投资人，甚至还有一位曾经的联合创始人。
如今，Cody已经加入Anthropic，Mithril也有了更多资源支持。
至于“大木博士”，目前还没有退休的打算。Cody表示，账号会继续更新，还是那些熟悉的话题，只是以后能够从AI药物研发的第一线分享更多故事。
在长文的最后，他化用了《宝可梦》系列最经典的宣传标语，只改了一个词：
“Gotta Cure 'Em All.”
从收服所有宝可梦，变成治愈所有疾病。
Anthropic也开始自己做药了
在亲自下场研发药物之前，Anthropic已经和不少大型药企建立了合作关系。
2025年10月，它推出Claude for Life Sciences，正式进军生命科学市场。当时，赛诺菲、诺和诺德等药企已经开始使用Claude，处理科研数据、药物开发文档和其他专业任务。今年1月，Anthropic又将产品扩展到临床试验管理、药品监管等环节。
Anthropic并不满足于只提供AI工具。今年4月，Anthropic以据报道约4亿美元的价格收购了生物医药AI公司Coefficient Bio，将一支专门研究AI药物研发的团队收入麾下。7月，它又买下了Mithril Medicine，也就是这次公布的“一人公司”。
8月27日，Anthropic开放了模型硬件标准MHS的研究预览，尝试让Claude直接协调和操作实验室设备。在与基因泰克的合作中，研究人员利用MHS连接自动移液器、机械臂和检测仪器，完成了一项蛋白质浓度检测的自动化实验。同一天，Anthropic还宣布向全球科研人员提供首批1万个免费或优惠订阅席位，并为符合条件的科研项目提供最高5万美元的Claude使用额度。
9月17日，Anthropic推出生命科学验证计划，允许经过资质、安全和伦理审核的研究机构，在调整后的生物安全限制下使用前沿模型，开展药物发现等研究。一天后，有外媒报道，Anthropic在旧金山湾区建立了自己的湿实验室，开始直接开展生物学研究。
昨天，Anthropic又正式公开了自己的药物研发项目。除了Mithril的两条研发管线，它还公布了与六家生物医药机构合作的早期成果。
其中，Anthropic与BigHat Biosciences合作，在约10周内制造并测试了约1010种抗体变体，部分优化后的分子与目标蛋白的结合能力提高了十倍以上。
另一家合作伙伴Enveda则利用Claude辅助审查了1061页药物生产文件，其中一次运行额外标记出7处人工审查未发现的问题。包含人工复核在内的审查时间从25.9小时缩短至2.8小时，减少约89%。
从向药企出售AI工具，到收购生物技术公司、建立自己的实验室，Anthropic如今也成为了药物研发的直接参与者。
只是，AI可以加快科研流程，却无法跳过漫长的临床试验。Anthropic已经开始自己做药，但距离真正把药送到患者手中，还有很长的路要走。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Letter AI
Anthropic announced a special deal yesterday: in July this year, it acquired a biomedical start-up company called Mithril Medicine. This company has only one full-time employee, while at the same time advancing a drug development project for two rare genetic diseases, one of which has already yielded preliminary results in mice.
The founders of the company, using Claude to coordinate 11 external laboratories, process experimental data, project arrangements and day-to-day operations, maintain the company alone.
It took less than 18 months between the development of the treatment concept and the completion of the initial animal experiment.
On the day the acquisition news was made public, Cody Tranbarger, the founder of Mistril, wrote a long article on X on publish:
“Surprise one: Anthropic acquired a single-person biotechnology company.”
" Surprise 2: That's me."
"Suspicion Three: I'm going to talk about it next..."
There's only one pharmaceutical company that Anthony sees.
Anthropic did not disclose the amount of the purchase, but made it clear that the two drug R & D pipelines of Mithril Medicine had become the first in-house drug R & D projects of the company.
In other words, Anthropic is personally involved in the development of new medicines, and this time the company it acquired has only one full-time employee.
Of course, a pharmaceutical company cannot do all the experiments on its own. Behind Mithril is the financial support of Atlas Venture, a part-time scientific consultant and 11 external laboratories for practical experiments.
The founder, Cody, was responsible for determining the direction of the research, formulating plans, judging the results of the experiments and coordinating cooperation between the different institutions, and, with regard to the heavy daily work, he gave it to Claude — which is probably one of the reasons why Anthropic saw the company.
According to research published by Anthropic, Claude has undertaken a large number of laboratory coordination, project handover, data collation and pilot planning, including even contract preparation, tracking of project progress and drafting of mail.
In the past, these tasks were usually carried out by specialized operators and scientists, and now Cody, with Claude's help, can manage research and development tasks that are scattered across different institutions. As a result, Mithril has been able to maintain a single full-time staff for the long term.
The two treatments developed by the company are directed at LACA2-related congenital mystic malnutrition and α-mysynthesis. Both rare genetic diseases destroy the connection between muscle cells and the surrounding tissue, causing severe muscle incapacitation and damage. Mithril tried to re-establish connections like bridges with a bipolar antibodies combined with muscle cells and surrounding support structures.
Among them, a preliminary animal experiment has been completed for LMA2 diseases. In mice receiving the highest dose of treatment, the holding and partial muscle force indicators are about twice as high as in untreated groups. During experimental observations, four out of eight mice were not treated, while only one in the highest dose group died.
For Anthropic, the attraction of Mithril may be this. It has already tried a special working model in a real drug research and development project: an experienced entrepreneur who, with Claude, coordinated external research resources to accomplish tasks that in the past required more internal staff.
Cody referred to the model as “Pipeline-in-a-Person” (one R&D pipe), and after the acquisition of Mistril, he continued to be in charge of the two original R&D lines and was preparing to launch more research projects on rare diseases.
According to the current plan, the fastest-growing projects are expected to enter human clinical trials by the end of 2028.
Gotta Cure 'Em All.
Over the past six years, Cody has been using the alias “Professor Oak” (Dr. Oki) on X to share research dynamics, investment perspectives and insights into the biomedical industry. Although he had become less anonymous over the years, he had never officially told everyone who he was.
It was only when Anthropic acquired it that he decided to reveal his identity.
He mocked himself in the text, and now X is so boring that he has to expose himself in person and give the timeline a little more pleasure.
Cody undergraduated at Amester College, United States, in biology and economics, and was also the goalkeeper of the school's long hockey team during university. After graduation, he first entered Goldman Sachs, where he worked in the financing and mergers of biomedicine enterprises, and then turned to investment in life sciences to work in Casdin Capital.
In 2019, Cody learned about the Harvard Business School MBA project by chance read the personal introduction of an alumni, Rich Holgan. Rich has set up a rare diseases pro bono organization called Cure Rare Disase, which works to develop individualized therapy for rare genetic patients, but the number of patients with some of these diseases is too small to attract traditional pharmacists into research and development.
Cody was thinking about it for the next few months, and finally sent a private letter to Rich through LinkedIn expressing his willingness to help.
The two were thus connected. Cody is gradually involved in the organization ' s consultancy work and is also helping to contact corporate sponsors. In 2020, he attended a charity event organized by Cure Rare Disease and later joined the organization ' s board of directors.
In an interview with the organization, Cody mentioned that she had been following a medical model called “n-of-1 medicine”, which was developing individualized treatment programmes for a very small number of patients, even single patients. He hoped that those patients would also have access to advanced treatment techniques and try to reduce the cost of developing such therapy.
Since then, Cody has joined the genetic editing company Prime Medicine, which has accumulated operational experience with more biotechnology enterprises.
Investment and entrepreneurship experiences were not always smooth, with Prime Medicine on the market in 2022, after which stock prices continued to decline. In an article published in 2025, Cody recalled that he ultimately failed to earn money beyond the full depth of the employee options he acquired in that year. However, he remains grateful for this experience, in which many former colleagues have been his mentors and friends to date.
In 2023, Cody joined Harvard University for a dual degree in biotechnology MS/MBA and participated in the creation of the biotechnology company Review Bio.
By 2025, with the help of the Atlas Venture team, he had founded Mithril Medicine.
In an identity announcement on 27 October, Cody specifically thanked Mike Gladstone, Jason Rhodes and others in Atlas Venture. He joked that he had made enough right decisions in the start-up process to convince them that Mithril and himself deserved a seed investment.
"After all, even broken clocks can be set twice a day." He said that he was grateful for having been able to obtain such support.
In the long run, he also thanked the Bio Twitter community, which he had mixed for many years. Cody stated that he had first dived on X for six years before he began posting as “Dr. Okaki”. Over the past decade or so, the platform has brought a great deal to him: industry knowledge, a great number of friends, many jobs, several mentors, some investors and even a former co-founder.
Today, Cody has joined Anthropic and Mithril has additional resources to support it.
As for Dr. Oki, there are no plans to retire. Cody indicated that the accounts would continue to be updated, or those familiar topics, but that more stories could be shared from the first line of AI drug development.
At the end of the long paper, he used the most classic slogan in the "Beauco Dreams" series and changed only one word:
“Gotta Cure 'Em All.”
From the acceptance of all treasures to the cure of all diseases.
Anthropic's already working on his own.
Anthropic has established a partnership with a number of large pharmaceutical companies before he/she ends up developing drugs in person.
In October 2025, it launched Claude for Life Science, officially on the life sciences market. At that time, companies such as Senofi, Noor and Nord had begun to use Claude to process scientific data, drug development files and other professional tasks. In January of this year, Anthropic expanded its products to include clinical trial management and drug regulation.
Anthropic is not content with providing only AI tools. In April of this year, Anthropic purchased Biomedicine AI Coefficent Bio at a reported price of about $400 million, under which a team dedicated to research and development of AI pharmaceuticals was brought together. In July, it bought again Mithril Medicine, the “one-man company” published on this occasion.
On 27 August, Anthropic opened a research preview of the model hardware standard MHS, attempting to direct Claude to coordinate and operate laboratory equipment. In cooperation with Gentek, researchers completed an automated test of protein concentration detection using the MHS connection to the ATM, mechanical arm and detection instrument. On the same day, Anthropic also announced the first 10,000 free or preferential subscription seats for global scientists and the use of quota by Claude for eligible research projects up to $50,000.
On 17 September, Anthropic launched the Life Sciences Certification Scheme, which allows research institutions with qualifications, safety and ethics clearance to use cutting-edge models and conduct research such as drug discovery under adapted biosafety restrictions. One day later, it was reported that Anthropic had established his own wet laboratory in the San Francisco Bay area and had begun direct biological research.
Yesterday, Anthropic officially announced his own drug research and development project. In addition to its two R & D pipelines in Mithril, it published early results in collaboration with six biomedical institutions.
Of these, Anthropic, in collaboration with BigHat Biosciences, manufactured and tested approximately 1010 antibody variants over a period of about 10 weeks, resulting in a more than tenfold increase in the ability of partially optimized molecules to combine with target proteins.
Another partner, Enveda, reviewed 1061 pages of pharmaceutical production documents with Claude ' s assistance, one of which operated an additional mark of seven manual reviews of undetected problems. The review time, including manual review, was reduced from 25.9 to 2.8 hours, a decrease of approximately 89 per cent.
From selling AI tools to pharmaceutical companies to acquiring biotechnology companies and establishing their own laboratories, Anthropic is now also a direct participant in drug research and development.
It is just that AI can speed up the scientific process but cannot skip long clinical trials. Anthropic has begun to do his own medicine, but there is still a long way to go before it is actually delivered to the patient.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.