GitLab is an open-source DevOps platform used by developers worldwide to plan, build, test, and deploy software — all in one application. Think of it as the control center for building software. It competes with GitHub (owned by Microsoft) but differentiates by being fully open-core and offering the entire DevOps lifecycle in a single tool.

After Y Combinator W15, GitLab grew to 2,500+ employees, went fully remote (no offices), and IPO'd on October 14, 2021. It reached a $6.4 billion market cap and generates roughly $800 million in annual revenue.

GitLab 是一个开源 DevOps 平台，全球开发者用它来规划、构建、测试和部署软件——全部在一个应用中完成。可以把它想象成构建软件的控制中心。它与 GitHub（微软旗下）竞争，但以完全开源核心和在单一工具中提供整个 DevOps 生命周期而著称。

经过 Y Combinator W15 之后，GitLab 发展到 2500+ 员工，完全远程办公（没有办公室），于 2021 年 10 月 14 日 IPO。市值达到 64 亿美元，年收入约 8 亿美元。

Sid later applied this same principle to his medical data. He published 25 terabytes of his diagnostic data on osteosarc.com — including single-cell sequencing, MRI scans, treatment timelines, and genetic analyses. This open-source approach to health data is unprecedented for an individual cancer patient, enabling researchers and other patients worldwide to learn from his journey.

Sid 后来将同样的原则应用于他的医疗数据。他在 osteosarc.com 上发布了 25 TB 的诊断数据——包括单细胞测序、MRI 扫描、治疗时间线和基因分析。这种开源的健康数据方式对个人癌症患者来说是前所未有的，使全球的研究人员和其他患者能够从他的历程中学习。

Osteosarcoma is a cancer that produces immature bone. It's the most common primary bone cancer, but still extremely rare — fewer than 1,000 new cases per year in the US. It typically strikes teenagers during growth spurts, making Sid's diagnosis at 45 highly unusual. The 5-year survival rate for localized disease is around 60-70%, but for recurrent or metastatic disease, outcomes are much grimmer.

骨肉瘤是一种产生不成熟骨骼的癌症。它是最常见的原发性骨癌，但仍然极为罕见——美国每年新发病例不到 1000 例。它通常在青少年生长发育期间发病，使得 Sid 在 45 岁时的诊断非常不寻常。局限性疾病的 5 年生存率约为 60-70%，但对于复发或转移性疾病，预后要严峻得多。

Spinal fusion surgery replaces damaged vertebrae with titanium hardware that permanently connects two or more vertebrae together. This stabilizes the spine but eliminates flexibility in that segment. For Sid, the surgery removed the T5 vertebra where the tumor was anchored and replaced it with metal scaffolding. Recovery from this surgery alone takes months.

SBRT (Stereotactic Body Radiation Therapy) delivers very high doses of radiation to a small, precisely defined area in just a few sessions (compared to weeks of conventional radiation). Proton beam therapy uses protons instead of X-rays, allowing more precise targeting that deposits most energy directly in the tumor while sparing surrounding tissue. Both represent the most advanced forms of radiation available.

脊柱融合手术用钛金属硬件替换损坏的椎骨，将两个或多个椎骨永久连接在一起。这稳定了脊柱但消除了该节段的灵活性。对于 Sid，手术切除了肿瘤所在的 T5 椎骨，并用金属支架替换。单是这个手术的恢复就需要数月时间。

Click chemistry refers to a set of simple, reliable chemical reactions where molecules "click" together like LEGO bricks. In cancer treatment, one component is injected near the tumor, and a drug-carrying component is injected into the bloodstream. The two click together only at the tumor site, concentrating the drug exactly where it's needed while sparing the rest of the body. Carolyn Bertozzi, Morten Meldal, and Barry Sharpless shared the 2022 Nobel Prize in Chemistry for developing this approach.

A single-patient IND (Form 3926) allows an individual patient to access an experimental drug outside of a clinical trial when no comparable treatment exists. The FDA approves 99.7% of these applications (they now say 100%) and typically responds within 48 hours. It's an underused pathway that most patients and even many oncologists don't know about. Sid has filed 5 such applications, all approved.

点击化学是指一组简单、可靠的化学反应，分子像乐高积木一样"咔嗒"拼接在一起。在癌症治疗中，一个组分注射到肿瘤附近，一个携带药物的组分注射到血液中。两者仅在肿瘤部位结合，将药物精确集中在需要的地方，同时保护身体其他部分。Carolyn Bertozzi、Morten Meldal 和 Barry Sharpless 因开发这种方法共享了 2022 年诺贝尔化学奖。

The medical system is designed around liability minimization, not survivability maximization. Doctors face IRB (Institutional Review Board) oversight, malpractice risk, and institutional inertia. Sid describes IRBs as a "vetocracy" — organizations whose power comes from saying no. The incentive structure means doctors default to the safest known path, even when that path leads nowhere for a patient with recurrent cancer. Sid's philosophy: "I'd rather die from a treatment than from the cancer."

医疗系统的设计围绕着最小化责任，而非最大化存活率。医生面临 IRB（机构审查委员会）监督、医疗事故风险和制度惯性。Sid 将 IRB 描述为"否决体制"——其权力来自于说不。激励结构意味着医生默认选择最安全的已知路径，即使对于复发癌症患者来说这条路已经走到了尽头。Sid 的哲学："我宁愿死于治疗，也不愿死于癌症。"

Traditional sequencing blends all cells together, giving you an average — like blending every instrument in an orchestra into a single note. Single-cell sequencing (using 10x Genomics technology) captures each cell individually, revealing the full diversity within a tumor. This exposed fibroblast markers (KERA, LUM, EPYC, FAP) in Sid's tumor that bulk sequencing would have averaged out. It was this single-cell data that ultimately led to the FAP-targeting radioligand therapy breakthrough.

传统测序将所有细胞混合在一起，给你一个平均值——就像将管弦乐队中的每种乐器混合成一个音符。单细胞测序（使用 10x Genomics 技术）单独捕获每个细胞，揭示肿瘤内的全部多样性。这暴露了 Sid 肿瘤中的成纤维细胞标记物（KERA、LUM、EPYC、FAP），这些在批量测序中会被平均掉。正是这些单细胞数据最终导致了 FAP 靶向放射配体治疗的突破。

The building blocks aren't prohibitively expensive: bulk RNA sequencing costs about $50, whole genome sequencing is around $500, and AI analysis tools run $20/month. What's expensive is the expertise to interpret the data, the access to the right tests, and knowing what to ask for. Sid's mission through osteosarc.com is to flatten this knowledge gap so future patients don't need to be tech billionaires to access frontier cancer diagnostics.

基本构建模块并不是高不可攀的：批量 RNA 测序大约 50 美元，全基因组测序大约 500 美元，AI 分析工具每月 20 美元。真正昂贵的是解读数据的专业知识、获取正确测试的途径，以及知道该问什么。Sid 通过 osteosarc.com 的使命是缩小这种知识差距，让未来的患者不需要成为科技亿万富翁才能获得前沿的癌症诊断。

Fibroblasts are the cells responsible for wound healing — they lay down collagen and structural tissue. FAP (Fibroblast Activation Protein) is normally found on fibroblasts during wound repair but disappears once healing is complete. In many tumors, FAP-expressing fibroblasts persist indefinitely, creating a suppressive cocoon that signals immune cells to stand down. By targeting FAP, you're not attacking the cancer cells directly — you're destroying the shield that protects them.

成纤维细胞是负责伤口愈合的细胞——它们铺设胶原蛋白和结构组织。FAP（成纤维细胞激活蛋白）通常在伤口修复期间出现在成纤维细胞上，但一旦愈合完成就会消失。在许多肿瘤中，表达 FAP 的成纤维细胞无限期存在，创造了一个抑制性茧，向免疫细胞发出停止信号。通过靶向 FAP，你不是直接攻击癌细胞——你是在摧毁保护它们的盾牌。

Radioligand therapy is like a guided missile system. A targeting molecule (the "ligand") seeks out a specific protein on cancer cells — in this case, FAP. Attached to the ligand is a radioactive isotope (Lutetium-177) that emits beta radiation, destroying nearby cells. The beauty: it delivers radiation precisely to the tumor while leaving most of the body untouched. The "theranostic" approach first uses a diagnostic version (cold scan) to confirm the target lights up, then switches to the therapeutic version (hot payload).

放射配体治疗就像一个导弹制导系统。靶向分子（"配体"）寻找癌细胞上的特定蛋白质——在这种情况下是 FAP。附着在配体上的是放射性同位素（镥-177），它发射β辐射，摧毁附近的细胞。其美妙之处在于：它将辐射精确递送到肿瘤，同时保持身体大部分不受影响。"诊疗一体化"方法首先使用诊断版本（冷扫描）确认靶标亮起，然后切换到治疗版本（热载荷）。

Unlike COVID mRNA vaccines that target a viral protein, cancer mRNA vaccines encode neoantigen peptides unique to a patient's tumor. Once injected, cells read the mRNA instructions and produce these neoantigens on their surface. The immune system recognizes them as foreign and mounts an attack — creating a targeted army of T-cells trained specifically against the patient's cancer. Moderna has shown that a personalized neoantigen vaccine plus checkpoint inhibitor cut melanoma recurrence/death risk nearly in half (December 2023 data).

与针对病毒蛋白的 COVID mRNA 疫苗不同，癌症 mRNA 疫苗编码患者肿瘤特有的新抗原肽。注射后，细胞读取 mRNA 指令并在其表面产生这些新抗原。免疫系统将它们识别为外来物质并发起攻击——创建一支专门针对患者癌症训练的 T 细胞大军。Moderna 已经表明，个性化新抗原疫苗加检查点抑制剂将黑色素瘤复发/死亡风险降低了近一半（2023 年 12 月数据）。

In computing, an AND-gate outputs TRUE only when all inputs are TRUE. In CAR-T therapy, the same concept is applied to biology: the engineered T-cell has two receptors. It only activates its killing function when BOTH receptors detect their targets simultaneously on the same cell. This dramatically reduces off-target toxicity — the #1 problem with conventional CAR-T therapy, which can cause fatal immune reactions when it attacks healthy tissue.

在计算中，AND 门只在所有输入都为 TRUE 时才输出 TRUE。在 CAR-T 治疗中，同样的概念应用于生物学：工程化的 T 细胞有两个受体。它只在两个受体同时在同一细胞上检测到目标时才激活杀伤功能。这大大减少了脱靶毒性——传统 CAR-T 治疗的第一大问题，当它攻击健康组织时可能导致致命的免疫反应。

PANX3 (Pannexin 3) is a membrane channel protein. Being hydrophobic means it repels water — and since most standard lab assays use water-based solutions, PANX3 simply doesn't show up. It's not in the GTEX database of normal tissue expression, meaning it's virtually absent in healthy tissue but massively overexpressed in Sid's tumor. This makes it a potentially ideal drug target with very low off-target risk — if you can develop a binder for it.

PANX3（Pannexin 3）是一种膜通道蛋白。疏水性意味着它排斥水——由于大多数标准实验室检测使用水基溶液，PANX3 根本不会出现。它不在正常组织表达的 GTEX 数据库中，这意味着它在健康组织中几乎不存在，但在 Sid 的肿瘤中大量过表达。这使它成为一个潜在的理想药物靶点，具有非常低的脱靶风险——如果你能为它开发一个结合剂的话。

Named as the reverse of Moore's Law (which describes exponential improvement in computing), Eroom's Law describes exponential DECLINE in pharmaceutical R&D efficiency. Despite massive advances in biology, genomics, and AI, the cost of bringing a new drug to market has been rising steadily since the 1950s. Reasons include: increasingly stringent regulation, the "low-hanging fruit" problem (easy diseases were cured first), larger clinical trials, and risk-averse institutional culture. Sid believes personalized medicine at scale can break this trend.

以摩尔定律的反向命名（摩尔定律描述计算能力的指数级提升），Eroom 定律描述了制药研发效率的指数级下降。尽管生物学、基因组学和 AI 取得了巨大进步，但将新药推向市场的成本自 1950 年代以来一直在稳步上升。原因包括：越来越严格的监管、"低垂果实"问题（容易的疾病先被治愈了）、更大的临床试验，以及规避风险的机构文化。Sid 相信大规模的个性化医学可以打破这一趋势。