12-Factor Agents: Patterns of reliable LLM applications — Dex Horthy, HumanLayer

AI Engineer · · Duration 17:06

A source by Dex Horthy of HumanLayer on agent reliability thresholds, developer ownership of generated code, and human-agent collaboration patterns in LLM application development. Read 3 viewpoints with supporting evidence and source links.

Understand this piece

3 key points

Synthesis

  1. 70–80% quality is insufficient for production

    Seventy to eighty percent functionality is enough to excite leadership and grow the team, but not enough to meet the reliability threshold for production use—beyond that point, developers confront deep, opaque call stacks and prompt or tooling complexity.

    Supporting evidence 1

    Original excerpt

    It's enough to get the CEO excited and get six more people added to your team. But then you kinda realize that seventy, eighty percent isn't quite good enough, and that if you wanna get past that seventy, eighty percent quality bar, you're seven layers deep in a call stack trying to reverse engineer how does this prompt get built

    Dex Horthy · 0:41

    Read in source context →
  2. Scaffolded, owned code over wrappers

    Dex Horthy argues against wrappers around internal abstractions and instead advocates for scaffolded, developer-owned code—comparing ideal tooling to shadcn, where generated code is fully owned and modifiable by the developer.

    Supporting evidence 1

    Original excerpt

    You don't need a wrapper around an internal thing. You need something more like shadcn, which is like scaffold it out, and then I'll own it, and I'll own the code and I'm okay with that.

    Dex Horthy · 15:00

    Read in source context →

    Continue exploring

    developer experience →
  3. Agents are better with humans

    Dex Horthy says agents are better with people and encourages builders to find ways for agents to collaborate with humans.

    Supporting evidence 1

    Original excerpt

    Um, and my take, agents are better with people. Find ways to let agents collaborate with humans.

    Dex Horthy · 15:38

    Context

    Um, there are hard things in building agents, but you should probably do them anyways, at least for now, and you should do most of them.

    Read in source context →

    Continue exploring

    human-agent interaction →

Key moments3

Short, attributed passages with the context to verify them. The full conversation stays with its publisher.

agent reliability threshold

70–80% quality is insufficient for production

Original excerpt

It's enough to get the CEO excited and get six more people added to your team. But then you kinda realize that seventy, eighty percent isn't quite good enough, and that if you wanna get past that seventy, eighty percent quality bar, you're seven layers deep in a call stack trying to reverse engineer how does this prompt get built
developer experience

Scaffolded, owned code over wrappers

Original excerpt

You don't need a wrapper around an internal thing. You need something more like shadcn, which is like scaffold it out, and then I'll own it, and I'll own the code and I'm okay with that.
human-agent interaction

Agents are better with humans

Original excerpt

Um, and my take, agents are better with people. Find ways to let agents collaborate with humans.
Context

Um, there are hard things in building agents, but you should probably do them anyways, at least for now, and you should do most of them.

Source & methodology

These viewpoints are linked to their original sources. Paraphrases are labeled and are not verbatim quotes.

Open transcript or source material (opens in a new tab)Report an issue

Explore these viewpoints by person