Carving the Ox: Agentic Data Ingestion and the Limits of Imitation

Alex Gaesser

Software Engineer

Parable

Software Engineer with deep experience building scalable platform services and APIs across multiple startups. Specialized in Go, distributed systems, and API design with strong emphasis on observability, testing, and operational excellence. Proven track record building shared libraries, establishing engineering standards, and delivering reliable systems serving thousands of enterprise customers..

Off-the-shelf ingestion sells the easy cuts: the same few hundred popular APIs everyone supports. The data you actually want sits in the long tail: APIs that are fully documented but that no tool bothers to build for. So we built agentic ingestion to carve the rest. Zhuangzi's master cook butchers an ox by working only the natural joints, and his blade stays sharp for nineteen years; a lesser cook hacks, and dulls his knife in a month. Our engine is the same: agents glide when they synthesize the patterns every API shares - auth, pagination, response patterns - and dull fast when they stop generalizing and copy one spec by rote. This talk shows where vibe-coding delivered, where it broke, and how to tell when "the AI will figure it out" is a trap.

Previous
Previous

Government Databases

Next
Next

Data Engineering Panel