From 57c42f4e1f447a82f819d9ac4b9565b5696a9b10 Mon Sep 17 00:00:00 2001 From: Richard Abrich Date: Sat, 17 Jan 2026 01:31:31 -0500 Subject: [PATCH] docs: qualify README claims for intellectual honesty - Change "Core Innovation" to "Core Approach" (more accurate) - Change "key differentiator" to "explores" (less marketing) - Correct accuracy figure (46.7% -> 100%, not 33% -> 100%) - Add context that all 45 tasks share same navigation entry point - Link to publication roadmap for methodology and limitations - Change "No technical expertise needed" to "Reduced prompt engineering" The goal is accuracy over marketing appeal. Co-Authored-By: Claude Sonnet 4.5 --- README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 8bffd41f5..0cd73e898 100644 --- a/README.md +++ b/README.md @@ -285,18 +285,18 @@ flowchart TB class L0,L1,L2 implemented ``` -### Core Innovation: Demo-Conditioned Prompting +### Core Approach: Demo-Conditioned Prompting -OpenAdapt's key differentiator is **demonstration-conditioned automation** - "show, don't tell": +OpenAdapt explores **demonstration-conditioned automation** - "show, don't tell": | Traditional Agent | OpenAdapt Agent | |-------------------|-----------------| | User writes prompts | User records demonstration | | Ambiguous instructions | Grounded in actual UI | -| Requires prompt engineering | No technical expertise needed | +| Requires prompt engineering | Reduced prompt engineering | | Context-free | Context from similar demos | -**Retrieval powers BOTH training AND evaluation**: Similar demonstrations are retrieved as context for the VLM, improving accuracy from 33% to 100% on first-action benchmarks. +**Retrieval powers BOTH training AND evaluation**: Similar demonstrations are retrieved as context for the VLM. In early experiments on a controlled macOS benchmark, this improved first-action accuracy from 46.7% to 100% - though all 45 tasks in that benchmark share the same navigation entry point. See the [publication roadmap](docs/publication-roadmap.md) for methodology and limitations. ### Key Concepts