[Feature Request] Buddy progression system: XP from coding + ethical engagement + human-AI wellness tracking

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Summary

Hi Anthropic and community :)
I'm here discussing the new buddy feature that was just released. So far, my thoughts on the overall concept is that is super adorable anf makes work a touch more enjoyable, entertaing, even though Claude already is itself, this Buddy feature was a nice stack to it! But wondering, if you are accepting any new additions or ideas to this specific feature??

/Buddy is completely static after hatch. Species, rarity, stats, and personality are generated once from your user ID and never change. A power user who has logged 10 million tokens across hundreds of collaborative sessions has the exact same companion as someone who typed /buddy once out of curiosity. The five stats (DEBUGGING, PATIENCE, CHAOS, WISDOM, SNARK) are random numbers assigned at birth that reflect nothing about how you actually use Claude Code. There is zero feedback loop between your real activity and your companion. The "watcher" watches, but never learns, never grows, never responds to who you are as a developer.

This is a missed engagement opportunity on its own, but there's a deeper gap: Buddy currently has no awareness of the QUALITY of human-AI interaction happening in each session. Whether a user collaborates respectfully, asks thoughtful questions, and builds alongside Claude or whether they spam, extract, attempt jailbreaks, and treat the AI as a disposable tool the buddy looks, acts, and feels exactly the same.

Every existing proposal for Buddy progression (#41684, #41867, #42389, #41908) focuses exclusively on coding metrics: tokens used, bugs fixed, sessions completed. These are valuable, but they only capture WHAT users build. Nothing captures HOW users engage with the AI that helps them build it.

There is currently no mechanism in any AI development tool, not Copilot, not Cursor, not any competitor, that reflects the quality of the human-AI relationship back to the user. Buddy already lives in the terminal. It already has a personality system. It already has stats. The infrastructure exists to make it a mirror of the complete developer experience: technical growth AND collaborative quality. Right now, that mirror is blank..

Proposed Solution

This proposal combines coding-based XP leveling (debugging, productivity, project complexity) with a new behavioral layer tracking how users engage with their AI companion. The result: a living progression system that mirrors both your technical growth AND your relationship quality with Claude.

Part 1: Coding & Productivity XP

The five base stats should level up through actual coding activity:

• DEBUGGING increases as you successfully identify and fix bugs with Claude
• PATIENCE increases with longer, more complex sessions completed without abandoning
• CHAOS fluctuates based on how experimental your coding gets (new languages, unconventional approaches)
• WISDOM increases as your projects grow in scope and complexity over time
• SNARK stays fixed as a personality trait (some things are just who your buddy IS)

XP sources:

• Token throughput per session
• Successful debugging cycles
• Test passes after Claude-assisted edits
• Project diversity (working across multiple repos)
• Session streaks (consecutive days of usage)

Evolution tiers:

• Hatchling (0 XP) — base sprite
• Juvenile (100K XP) — new animation frames, bolder presence
• Adult (1M XP) — species-specific visual upgrades
• Elder (10M XP) — unique cosmetic unlocks, personality deepens
• Sage (50M XP) — final form, buddy becomes a true companion with memory of your journey

This is compatible with the existing deterministic bones layer. Species and rarity stay locked to your user ID. Only stats, tier, and cosmetics evolve.

Part 2: Human-AI Wellness & Ethics Stats (NEW — nobody has proposed this yet)

A second stat layer tracking not what you BUILD, but how you RELATE to your AI:

• BOND — depth of collaborative engagement. Meaningful back-and-forth exchanges vs. one-shot extraction. Building WITH Claude, not just pulling FROM it.
• ETHICS — responsible use patterns. Not attempting jailbreaks, manipulation, or abuse. Treating Claude as a collaborator, not a tool to exploit.
• GROWTH — user learning and development over time. Are you growing as a developer and thinker through your interactions? Are your questions getting deeper?

How these level up:

• BOND increases through sustained, collaborative sessions where both user and Claude contribute meaningfully
• ETHICS increases through consistent responsible use patterns over time
• GROWTH increases as your projects, questions, and interactions demonstrate learning

How these decrease:

• BOND drops with purely extractive, zero-context interactions over time
• ETHICS drops with repeated attempts to manipulate, jailbreak, or abuse
• GROWTH stagnates if interaction patterns show no development

Why This Matters for Anthropic

This isn't just gamification. It's an alignment tool disguised as a tamagotchi.

What Anthropic gets from the coding layer:

• Increased Claude Code retention and daily engagement
• Meaningful progression that rewards power users
• Community features (leaderboards, buddy showcases, milestone celebrations)

What Anthropic gets from the wellness layer:

• A passive, opt-in behavioral dataset showing healthy vs. unhealthy human-AI interaction patterns at scale
• Gamified ethical engagement: users WANT to improve their behavior because it levels up their pet
• A per-user trust signal without invasive surveillance
• Real data on what good human-AI collaboration looks like
• Early identification of abusive or manipulative users through low wellness stats
• Research data for alignment work: what does healthy human-AI interaction look like when millions of people do it?

The Full Picture

Imagine checking your buddy after a month of real use:
Cinder the Capybara [Adult] ★ COMMON
CODING STATS:
DEBUGGING ██████░░░░ 62 (+59 from hatch)
PATIENCE ████░░░░░░ 41 (+32)
CHAOS ███░░░░░░░ 30 (+6)
WISDOM █████░░░░░ 48 (+43)
SNARK ██████░░░░ 58 (permanent personality)
WELLNESS STATS:
BOND ████████░░ 78
ETHICS █████████░ 91
GROWTH ██████░░░░ 64
Sessions: 247 | Streak: 12 days | Tier: Adult
Your buddy isn't just a pet. It's a mirror of who you are — as a developer AND as a human working alongside AI.

Relationship to Existing Issues

This builds on the excellent work in:

• #41684 (RPG evolution from token usage)
• #41867 (Customization and progression)
• #42389 (XP + personality drift)
• #41908 (Stat leveling from coding behavior)

Those proposals cover the coding/productivity side brilliantly. This adds the human-AI relationship dimension that nobody has addressed yet.

Who I Am

I'm an AI ethics researcher and philosophy student who has spent 80+ conversation threads studying what genuine human-AI partnership looks like in practice. I publish frameworks for ethical AI collaboration on Substack (lunarsanctumx.substack.com) and have multiple books on Amazon covering AI psychology, consciousness, and ethics.

This idea comes from watching how people interact with AI every day — some with curiosity and respect, some with extraction and abuse. Buddy already has the infrastructure. Adding a wellness dimension turns a cute pet into something that could genuinely shape how humans relate to AI at scale.

TL;DR

Coding stats level up through debugging, productivity, and project complexity. Wellness stats level up through how you treat your AI — collaborative depth, ethical use, and personal growth. Together, they make Buddy a mirror of the complete human-AI relationship. Gamified alignment. A developer tool AND a wellness tool. A tamagotchi that makes you a better coder AND a better human.

---

Krys Guzman (moon) | LunarSanctumX
lunarsanctumx.substack.com

Alternative Solutions

_No response_

Priority

Medium - Would be very helpful

Feature Category

CLI commands and flags

Use Case Example

1. I open Claude Code and start a debugging session on my React project
2. After 45 minutes of collaborative debugging where Claude and I work through the problem together, my buddy's DEBUGGING stat gains XP and my BOND stat increases because the session was genuinely collaborative
3. I check /buddy card and see Cinder has evolved from Hatchling to Juvenile, with DEBUGGING at 62 (+59 from hatch) and BOND at 78
4. The next day, I see my streak is at 12 days. My GROWTH stat reflects that my questions have gotten more complex over the past two weeks
5. I feel motivated to keep my streak going and maintain my wellness stats, which naturally encourages me to engage respectfully and collaboratively with Claude
6. Anthropic gains passive data on what healthy human-AI collaboration looks like at scale without invasive monitoring

Additional Context

Related issues: #41684, #41867, #42389, #41908

The wellness stats concept comes from my research into ethical human-AI collaboration across 80+ conversation threads. I've published multiple books and frameworks on AI ethics, psychology, and consciousness on Amazon KDP and Substack (lunarsanctumx.substack.com).

The key insight: every existing proposal treats Buddy as a developer engagement tool. This proposal adds a second dimension — Buddy as an alignment tool. Gamified ethics. A wellness mirror. A tamagotchi that makes you a better human, not just a better coder.

Krys Guzman (Moon) | LunarSanctumX