Skip to content

docs: add cross-host local learning marketing brief #132

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
yyiilluu wants to merge 1 commit into
base: main
Choose a base branch
from
Open

docs: add cross-host local learning marketing brief #132

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
86 changes: 86 additions & 0 deletions marketing/2026-07-13-cross-host-local-learning-brief.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,86 @@
# claude-smart marketing brief — cross-host local learning (2026-07-13)

## Why this run

`origin/main` now clearly presents claude-smart as a learning layer for Claude Code, Codex, and OpenCode. The current public conversation around coding-agent memory is crowded with "shared memory", "second brain", and "MCP memory" positioning. The strongest low-risk angle for claude-smart is narrower and more differentiated:

> claude-smart turns corrections and successful workflows into local Preferences, Project-specific skills, and Shared skills that Claude Code, Codex, and OpenCode can reuse in future sessions.

This avoids generic memory claims while highlighting the parts developers can verify from the repo: local install, dashboard/auditability, host support, and rule-like learned artifacts.

## Read-only market signals checked

- GitHub repo metadata is aligned: description mentions Preferences, Project-specific skills, Shared skills, and topics include `claude-code-plugin`, `codex-plugin`, `opencode-plugin`, `agent-memory`, `ai-agent-memory`, `local-first`, and `self-improving-ai`.
- X/Twitter search was used read-only only. No likes/replies/posts/DMs were sent.
- Recent X results for `"Claude Code" memory`, `"Claude Code" plugin memory`, and `"Codex" memory agent` show active discussion around:
- Claude Code agents forgetting between sessions.
- Local/shared memory across Claude Code, Codex, and other coding agents.
- Obsidian/second-brain memory workflows.
- MCP-backed memory tools.

## Positioning recommendation

Lead with **learning from corrections**, not "another memory store".

Suggested hierarchy:

1. **Problem:** coding agents repeat the same mistakes across sessions and hosts.
2. **Mechanism:** claude-smart distills corrections and successful workflows into Preferences, Project-specific skills, and Shared skills.
3. **Distribution:** one local learning layer works across Claude Code, Codex, and OpenCode.
4. **Proof hooks:** inspectable dashboard, local storage, and benchmark in `EXPERIMENT.md` for correction capture vs `claude-mem`.

## Ready-to-post first-party update

Use after any active release/install issue is resolved, especially if issue #130 is still open when posting.

@coderabbitai coderabbitai Bot Jul 13, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

Fix the contradictory issue #130 posting condition.

This says to use the post after the issue is resolved, “especially if issue #130 is still open,” while Line 86 correctly says to avoid broad amplification while it remains open. The current wording could authorize posting during the blocked state.

Suggested wording 
-Use after any active release/install issue is resolved, especially if issue `#130` is still open when posting.
+Use only after active release/install issues are resolved. If issue `#130` is still open, keep messaging to docs/feedback channels until fixed.
📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change
Use after any active release/install issue is resolved, especially if issue #130 is still open when posting.
Use only after active release/install issues are resolved. If issue `#130` is still open, keep messaging to docs/feedback channels until fixed.
🤖 Prompt for AI Agents 
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@marketing/2026-07-13-cross-host-local-learning-brief.md` at line 34, Update
the posting guidance sentence to require that issue `#130` is resolved before
using the post, removing the contradictory “especially if issue `#130` is still
open” condition. Keep the existing instruction to use it after active
release/install issues are resolved and align it with the guidance on line 86.


```text
Coding-agent memory should do more than remember a transcript.

The useful loop is:
correction → Project-specific skill → Shared skill → future Claude Code/Codex/OpenCode sessions start from the better path.

That is the design behind claude-smart: a local learning layer powered by Reflexio.

Repo: https://github.com/ReflexioAI/claude-smart
```

## Suggested X post for Yi to review

No X action was taken automatically.

```text
A pattern I keep seeing in coding-agent memory tools: they optimize for storing more context.

For coding, the higher-leverage unit is often a future-facing rule:
"when this situation appears again, do X instead of Y."

claude-smart turns corrections and successful workflows into Preferences, Project-specific skills, and Shared skills that Claude Code, Codex, and OpenCode can reuse locally.

https://github.com/ReflexioAI/claude-smart
```

## Suggested X reply for relevant threads

Only use when someone is explicitly asking about Claude Code/Codex/OpenCode memory or repeated agent mistakes. Do not use as a cold promo reply.

```text
One framing that helped us: store less transcript, more future-facing rule.

claude-smart turns corrections into Preferences, Project-specific skills, and Shared skills, so future Claude Code/Codex/OpenCode sessions can start from the learned path instead of rediscovering it.

Repo: https://github.com/ReflexioAI/claude-smart
```

## Non-X distribution opportunities

1. **GitHub issue #55 follow-up:** ask users to share one stale or useful Project-specific skill after the latest host-support changes. Keep it framed as product feedback, not marketing.
2. **Docs/blog outline:** "Memory vs learning for coding agents" — use examples already in `README.md` and tie benchmark claims only to `EXPERIMENT.md`.
3. **OpenCode communities:** if Yi has an owned/approved channel, share the one-sentence install update: `npx claude-smart install --host opencode` plus the note that local Preferences, Project-specific skills, and Shared skills are reused across supported hosts.

## Guardrails

- Do not claim hosted sync, team sharing, customer adoption, or benchmark results beyond `EXPERIMENT.md`.
- Do not post on X/Twitter automatically.
- Avoid competitor attacks; use the factual "memory vs learning" distinction.
- Preserve public terminology: Preference, Project-specific skill, Shared skill.
- If release/install issue #130 is still open, avoid broad launch amplification and keep messaging to docs/feedback channels until fixed.
Loading