WHAT NEW IN X10.31 SAT

X 上的 AI 合集档案 · 22 个合集 · 63,170 条

jevIMAGE

Jev 工作流把 AI 客服成本砍 90%

TypeSafe CEO 分享的 Jev 客服流程:先查缓存结果,再决定用预设回复、Opus 调查还是转人工,并记录决策日志。

Branko Trcek
@Brankotrcek19 天前收录原推发布于 2026.10.12
Jev 工作流把 AI 客服成本砍 90%
Branko Trcek @Brankotrcek原推发布于 2026.10.12
在 X 上打开原推

THIS IS F*CKING INSANE Diogo Almeida, CEO of TypeSafe AI, just shared a Jev workflow that cuts AI support costs by 90%. pair it with Opus 5.5 and stop overpaying for routine tickets time required: 1 minute action: copy and paste into your AI agent the budget killer is sending every ticket to the strongest model, then paying again when it retries without knowing what failed discovery before code: have the agent ask 7 questions about your support process, approved answers, escalation rules, and what counts as a resolved ticket Jev routing: check saved results first, then decide whether to use an approved response, launch an Opus investigation, or request human review decision log: record the route, the reason, and the outcome. see which calls were useful and which tickets the filter wrongly rejected targeted fixes: send Opus the failed check and relevant evidence. one correction instead of another full run with the same instructions an article below takes this further: saved results, fewer retries, and budget calculations for a workflow targeting 20x efficiency ↓

这条是引用着另一条写的:x.com/0xnineinch/status/2108604801411879139

x.com/brankotrcek/status/2109419442643845309采集于 2026.10.12 10:05