用专用if-then模型替代通用LLM，成本降百倍且准确率翻倍
AI Comes for the If Statement
给出了具体的技术选型方案、性能对比数据和成本数字，独立开发者和产品团队可直接用于优化现有AI功能的成本结构。
Working in the back of a grocery store has its own set of rules : if the food is a banana, send it to produce ; if it is a cookie, the snack aisle ; if it is cumin, shelve it with the spices.
在杂货店后仓工作有一套自己的规则：如果是香蕉，就送到果蔬区；如果是饼干，就放到零食货架；如果是孜然，就和香料放在一起。
But what if the load of bananas has spoiled, the cookies have crumbled & the cumin is caked? These rules & exceptions govern every grocery store & neighborhood mart. At the beginning they are rigid ; over time, more exceptions are discovered : is that a plantain? Dubai chocolate : dessert or baking supply?
但如果那批香蕉坏了，饼干碎成了渣，而孜然结块了怎么办？这些规则和例外情况适用于每家杂货店和社区便利店。起初它们很死板；随着时间的推移，人们发现了更多的例外情况：那是芭蕉吗？迪拜巧克力：算作甜点还是烘焙原料？
Coding literally means encoding these rules & exceptions into software. Pre-AI, these programs were rigid. AI handles the exceptions : an image search identifies the unfamiliar fruit as Musa paradisiaca, a brother of the banana.1
编程字面意思就是将这些规则和例外编码到软件中。在 AI 出现之前，这些程序是僵化的。AI 能够处理例外情况：图像搜索将这种陌生的水果识别为 Musa paradisiaca（芭蕉），它是香蕉的兄弟。1
But we do not need the world’s most brilliant model to handle if-plantain-then-produce logic.
但我们不需要世界上最强大的模型来处理“如果是芭蕉则放入果蔬区”的逻辑。
The newest wave of AI is a robust if-then decider. Jev2 & SemIf3 answer questions like these in hundreds of milliseconds at a 99% reduction in cost compared to traditional AI.4 They simplify the existing models by running the attention math once & then determining the probability of each allowed answer, based on a few tokens of output.5
最新一代 AI 是一种稳健的 if-then（如果-那么）决策器。Jev2 和 SemIf3 能在几百毫秒内回答这类问题，与传统 AI 相比成本降低了 99%。4 它们通过仅运行一次注意力机制计算，然后根据少量输出 token 确定每个允许答案的概率，从而简化现有模型。5
I went looking for the if-then statements in my own code, the ones I had handed to AI. Within a few minutes I had replaced about a quarter of those calls in one of my agents. On 98 hand-verified production email threads, the specialized deciders nearly doubled the classification accuracy of the production model, jumping from 47% to over 80%.6
我在自己的代码中寻找那些我交给 AI 处理的 if-then 语句。短短几分钟内，我就在一个智能体中替换了约四分之一的调用。在 98 个经过人工验证的生产环境邮件线程上，专用决策器几乎使生产模型的分类准确率翻了一番，从 47% 跃升至 80% 以上。6
Software is composed of primitives : the if-then statement is one of them. By optimizing this single primitive, we see nearly two orders of magnitude in cost reduction alongside higher accuracy.
软件由基本原语构成：if-then 语句就是其中之一。通过优化这一单一原语，我们实现了近两个数量级的成本降低，同时提高了准确性。
These advances raise the question of which other programming primitives will benefit from the same specialization. They also highlight the bifurcating economics of AI : state of the art models for discovery & optimized models for production. We use the largest frontier models to train new models, & the most capable models to architect systems. But once a system is engineered & hardened, running it thousands or millions of times through a workflow benefits from narrower AI.
这些进展引发了一个问题：还有哪些其他编程原语将从同样的专业化中受益。这也凸显了 AI 经济学的分化趋势：用于探索的最先进模型与用于生产的优化模型并存。我们使用最大的前沿模型来训练新模型，并使用最强大的模型来设计系统架构。但一旦系统经过工程化设计和加固，在成千上万次的工作流中运行它，采用更窄范围的 AI 会更有益。
If this is the first of many primitives specialized for production, then harnesses are about to capture a lot more margin.7
如果这是众多专为生产优化的原语中的第一个，那么Harnesses 即将捕获更多的利润空间。7
- Musa × paradisiaca. Linnaeus named plantains Musa paradisiaca & dessert bananas Musa sapientum in 1753 ; both are now known to be hybrids of the same two parents, Musa acuminata & Musa balbisiana. The classification was hard enough that he split one group into two. ↩︎
- Jev, TypeSafe, released 15 September 2026. Returns typed, calibrated decisions in one parallel pass rather than generated text. ↩︎
- SemIf, Theodore Lee, MIT licensed, released 18 September 2026, & kev, Jared Palmer : open reproductions of the interface pattern. Benchmarks are SemIf’s published figures on an RTX 3090 with a frozen Qwen3.5-4B. ↩︎
- Jev pricing : $0.042 per million input tokens, $0 output, which TypeSafe says may be subsidized. Sonnet-class pricing is $3 input & $15 output per million. On a typical classification call of 2,000 input & 60 output tokens that is 82 times cheaper. TypeSafe’s own workflow evals put Jev at $0.0004 a case against $0.0304 & $0.0836 for two frontier models, a 76x to 209x spread. ↩︎
- SemIf mechanics. Rather than generating tokens autoregressively through the full transformer stack, the open source implementation runs the attention mechanism over the prompt and evaluates candidate choices directly from the output logits, skipping the multi-layer feed-forward network and decode steps entirely. ↩︎
- Evaluation on 98 hand-verified production email threads, September 2026 : across the exact same dataset, the production generative LLM classifier achieved 47% accuracy (46/98), while Jev scored 80% (78/98) & local SemIf scored 82% (80/98). In addition, in live production logs across 31 inbound emails, the local decider acted on 8 with zero errors & safely deferred the rest to the frontier model. ↩︎
- The Harness Margin Opportunity. A harness cuts execution cost without sacrificing accuracy ; specializing the underlying model primitives expands that gross margin wedge even further. ↩︎
- Musa × paradisiaca（芭蕉）。林奈于1753年将植物蕉命名为 Musa paradisiaca，将食用香蕉命名为 Musa sapientum；现已知二者均为同一双亲本 Musa acuminata 与 Musa balbisiana 的杂交种。分类之难足以使他将该组一分为二。↩︎
- Jev，TypeSafe，2026年9月15日发布。在一次并行传递中返回经过类型检查和校准的决策，而非生成文本。↩︎
- SemIf，Theodore Lee，MIT 许可，2026年9月18日发布，以及 kev，Jared Palmer：接口模式的开源复现。基准测试数据为 SemIf 在 RTX 3090 上使用冻结版 Qwen3.5-4B 发布的指标。↩︎
- Jev 定价：每百万输入 token 0.042 美元，输出 token 0 美元，TypeSafe 称这可能由补贴支持。Sonnet 级别的定价为每百万输入 3 美元、输出 15 美元。对于典型的包含 2,000 个输入 token 和 60 个输出 token 的分类调用，其成本低 82 倍。TypeSafe 自身的工作流评估显示，Jev 每案例成本为 0.0004 美元，而两款前沿模型分别为 0.0304 美元和 0.0836 美元，差距达 76 倍至 209 倍。↩︎
- SemIf 机制。开源实现并未通过完整的 Transformer 堆栈自回归地生成 token，而是在提示词上运行注意力机制，并直接从输出 logits 评估候选选项，完全跳过多层前馈网络和解码步骤。↩︎
- 2026年9月在98条人工验证的生产邮件线程上的评估结果：在完全相同的数据集上，生产环境中的生成式 LLM 分类器准确率为 47%（46/98），而 Jev 得分为 80%（78/98），本地 SemIf 得分为 82%（80/98）。此外，在涵盖 31 封入站邮件的真实生产日志中，本地决策器对其中 8 封邮件的处理零错误，并将其余邮件安全地交由前沿模型处理。↩︎
- Harness 的利润空间机会。Harness 在不牺牲准确性的前提下降低了执行成本；通过专业化底层模型原语，进一步扩大了毛利空间。↩︎
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力