# Kuken Industries searches 70 years of drawings in minutes

Source: [OpenAI](https://openai.com/index/kuken)  
Feed7 permalink: https://feed7.dev/p/kuken-02ffah9  
Published: 2026-11-09T08:00:00.000Z  
Trust: Official Source (official_source)

## Why Included

Kuken’s drawing search cuts the reported time to find a repair drawing from nearly an hour to 4.5 minutes. The app also lets field engineers search with part numbers or drawing images.

## Source Summary

Kuken and Recursive built an app over a **70-year** archive of cooling tower drawings. Engineers can search by number or upload a drawing image or PDF; Kuken reports a drawing lookup now takes **4.5 minutes**, down from nearly an hour.

## Practical Implication

For agent builders, the useful pattern is combining retrieval with domain-specific checks: the app returns possible matches, then engineers compare tower specifications and replacement records. Kuken also says easier access lets newer engineers handle more work independently.

## Agent-Ready Context

Kuken and Recursive built an app over a **70-year** archive of cooling tower drawings. Engineers can search by number or upload a drawing image or PDF; Kuken reports a drawing lookup now takes **4.5 minutes**, down from nearly an hour.

For agent builders, the useful pattern is combining retrieval with domain-specific checks: the app returns possible matches, then engineers compare tower specifications and replacement records. Kuken also says easier access lets newer engineers handle more work independently.

The reported time saving comes from Kuken’s use and testing of the tool. A separate drawing-checking tool is still in development; its prototype flagged genuine discrepancies about 90% of the time, according to Kuken.

## Connected Context

Feed7 judgment across 1045 accumulated Signals:

This deployment illustrates retrieval over a specialized archive paired with domain checks by engineers, and reports faster lookup in Kuken’s own use. It grounds retrieval choices in a concrete image-and-document search task, while the separate prototype checker and company-reported results limit what can be inferred about automated validation or general performance.

- [Stop Chunking Like It's 2022 — Yuval Belfer, AI21 Labs](https://feed7.dev/p/stop-chunking-like-it-s-2022-yuval-belfer-ai21-labs-1bsztja) — Both concern retrieval over documents; Kuken supplies a concrete archive and query task against which chunking and indexing strategies could be evaluated.
- [AI on Your Lakehouse: Context Comes in Shapes, Not Queries — Zach Blumenfeld, Neo4j](https://feed7.dev/p/ai-on-your-lakehouse-context-comes-in-shapes-not-queries-zach-blumenfeld-0r4s9u5) — Kuken’s engineers compare candidate drawings with specifications and replacement records, illustrating the value of domain relationships alongside document retrieval.
- [Citation Needed: Provenance for LLM-Built Knowledge Graphs — Daniel Chalef, Zep AI](https://feed7.dev/p/citation-needed-provenance-for-llm-built-knowledge-graphs-daniel-chalef-1iob5t8) — Kuken reports engineers checking possible matches against domain records, reinforcing the distinction between retrieval assistance and human validation.

## Context Map

- Layer: context
- Domains: data
- Topics: retrieval

## Uncertainty

- The reported time saving comes from Kuken’s use and testing of the tool. A separate drawing-checking tool is still in development; its prototype flagged genuine discrepancies about 90% of the time, according to Kuken.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.
