From 1ba30706ef67cff412145614cfc8235ce577dc26 Mon Sep 17 00:00:00 2001 From: guangyu-reflexio Date: Tue, 25 Aug 2026 14:59:12 -0700 Subject: [PATCH 1/2] docs: call Reflexio a learning platform, not a layer The team standardised on "learning platform" as the product category. This updates the two places in this repo that name it something else: - AI_AGENT_INTEGRATION.md: "best-effort learning layer" -> "best-effort learning platform" - benchmark/gdpval/RESULTS.md: "an external self-improvement layer" and "Reflexio, a layer that reads a completed run" -> "learning platform" Phrases describing the *agent* as self-improving are deliberately left alone: "an agent that is already self-improving" is a claim about the customer's agent, not about Reflexio's category. --- AI_AGENT_INTEGRATION.md | 2 +- benchmark/gdpval/RESULTS.md | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/AI_AGENT_INTEGRATION.md b/AI_AGENT_INTEGRATION.md index e9538ec42..daac6cbfb 100644 --- a/AI_AGENT_INTEGRATION.md +++ b/AI_AGENT_INTEGRATION.md @@ -162,7 +162,7 @@ REFLEXIO_API_KEY="..." ``` Implementation rule: if Reflexio is unavailable, the agent must continue -normally. Treat Reflexio as a best-effort learning layer, not as a dependency +normally. Treat Reflexio as a best-effort learning platform, not as a dependency that can break user work. ## Configure an LLM Provider diff --git a/benchmark/gdpval/RESULTS.md b/benchmark/gdpval/RESULTS.md index ab5acd31b..8653c1db6 100644 --- a/benchmark/gdpval/RESULTS.md +++ b/benchmark/gdpval/RESULTS.md @@ -5,10 +5,10 @@ Modern AI agents already try to improve themselves between runs: they cache skills, remember prior plans, and get faster on tasks they have seen before. This raises a pointed question: **is there still room -for an external self-improvement layer to add measurable value on top +for an external learning platform to add measurable value on top of an agent that is already self-improving?** -We test **Reflexio**, a layer that reads a completed run of a task +We test **Reflexio**, a learning platform that reads a completed run of a task and produces a concrete, copy-pasteable recipe for the next run of the same task, on **five real knowledge-work tasks** from the public OpenAI GDPVal benchmark. These five tasks were selected because the From ccfbf64e72c1e80aad16e16007afd24ffc699298 Mon Sep 17 00:00:00 2001 From: guangyu-reflexio Date: Tue, 25 Aug 2026 15:16:53 -0700 Subject: [PATCH 2/2] docs: fix the SDK README's category claim MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit client_dist/README.md called Reflexio "the adaptive memory layer for AI agents". Two problems: 1. It is the wrong category term after the standardisation. 2. It directly contradicts the product's own positioning — the homepage leads with "Not another memory layer." and an entire blog post (remembers-everything-learns-nothing) argues that remembering is not learning. This file is the `readme` in client_dist/pyproject.toml, so it is the PyPI long-description for reflexio-ai: the most-read category statement Reflexio publishes outside the website. Found by review while checking the terminology sweep for completeness. --- client_dist/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/client_dist/README.md b/client_dist/README.md index 192a24f3f..76693ad40 100644 --- a/client_dist/README.md +++ b/client_dist/README.md @@ -5,7 +5,7 @@ [![License: Apache 2.0](https://img.shields.io/badge/license-Apache%202.0-green)](https://github.com/ReflexioAI/reflexio/blob/main/LICENSE) [![Downloads](https://static.pepy.tech/badge/reflexio-ai/month)](https://pepy.tech/project/reflexio-ai) -The official Python SDK for [Reflexio](https://www.reflexio.ai/) — the adaptive memory layer for AI agents. Reflexio automatically extracts user profiles, generates playbooks, and evaluates agent performance from conversation data. This client provides type-safe, sync-first access to the full Reflexio API. For source code and contributions, see the [GitHub repository](https://github.com/ReflexioAI/reflexio). +The official Python SDK for [Reflexio](https://www.reflexio.ai/) — the learning platform for AI agents. Reflexio automatically extracts user profiles, generates playbooks, and evaluates agent performance from conversation data. This client provides type-safe, sync-first access to the full Reflexio API. For source code and contributions, see the [GitHub repository](https://github.com/ReflexioAI/reflexio). ## Table of Contents