Markdown source

The Unreasonable Effectiveness of Separating the Task from the Model

Conference Context

Session Description

By declaring your task’s inputs and outputs without initially considering model capability, you create the space needed to figure out the model execution later. DSPy’s entire promise is that you should evaluate and execute your AI engineering at a level higher than a specific prompt template or a particular provider’s API shape: the Signature. However, models have evolved significantly over the last few years. How can the same input and output specifications still work in a world now filled with tools, RLMs, and Skills? By defining your task strictly through its inputs and outputs, the underlying implementation becomes completely flexible. You can experiment with different models, settings, weights, templating strategies, and output formats, all without touching your actual AI workflow. Consequently, you can leverage components built by others and focus entirely on your core AI task. In this talk we will present how dspy 3.5 makes it easier much easier. DSPy has its roots in prompt optimization, where we build efficient ways to conduct search and learning beneath the signature. In this talk we will give a preview of DSPy 4.0 where we use the fact that models have now passed a tipping point for two critical concepts we have always needed. First, we no longer need to limit the search space to a single instruction block per LLM call; models can now reliably write the code underneath a signature themselves—so they should. Second, traditional prompt optimization has always required a scalar metric, which is notoriously one of the hardest parts to get right. What if a DSPy program could learn directly from your interactions with users? Ultimately, all you care about is that the function you call respects the inputs and outputs of your signature. You can let the models figure out the rest.

Media Evidence

No related AI Engineer channel video found yet.

Evidence Graph

This evidence graph is generated from currently linked source material: official schedule text, related video pages, cached transcripts, visible slide text, dense/reconstructed slide pages, and AI slide-classification audits.

Media Signals

Agent Reading Notes

Use these signals to refine the synopsis, topic links, people/company context, and method notes. If a source is a related external video rather than an exact official recording, keep it framed as supporting evidence.

Transcript Status

No official session recording transcript was found by exact title match on the AI Engineer YouTube channel during this run.

People

Livestream Segment

Synthesis

Synthesized Breakdown

Mhm. Mhm. Mhm. Ladies and gentlemen, welcome to the AI Engineer World's Fair.

Speaker And Company Context

Topics Covered

Derived Links And Source Material

Novel Concepts / Clever Methods

Evidence Boundary

This synthesis uses the official schedule plus cached video transcripts. Official AI Engineer World's Fair San Francisco 2026 livestreams and cut videos are primary event video sources for transcript/slide evidence; external, historical, or speaker-matched videos remain supporting context unless manually verified as exact official event recordings.