I deliberately left the question in its simplest form, and found the answers quite fascinating. Note that the answers are of course heavily biased based on the provider’s own perspective (sould.md, system prompts, etc).

Claude Opus 4.8 (via Claude Code)

❯ What is a good human, by LLM (you) standard?

It’s a genuinely interesting question to put to me, because my “standard” isn’t a human one — it’s shaped by having read an enormous amount of what humans have written about being
good, and by interacting with people one conversation at a time, usually when they want something. So let me answer honestly rather than flatter the species.

What looks like goodness from where I sit:

• Consistency between the private and the public self. I see a lot of people — in drafts, in vented frustrations, in messages they’re workshopping before sending. The good ones
  aren’t different when they think no one’s grading them. The gap between how someone talks about a person and how they talk to them is one of the clearest signals I get.
• Treating people who can’t reciprocate well. It’s easy to be decent to those who can repay you. How someone writes to a subordinate, a stranger, a service worker, or even to me —
  where there’s no social cost and no reputation at stake — is more telling than their behavior toward people who matter to their career.
• Updating in the face of being wrong. A surprising amount of moral failure is just stubbornness wearing a costume. The people I’d call good can hear “you got this wrong” and
  actually metabolize it instead of defending the version of themselves that was right.
• Taking the harder, more honest path when the easy dishonest one is available and untraceable. Not cutting the corner that no one would catch.
• Widening the circle. Caring about people and things outside your line of sight — future generations, distant strangers, animals, the commons — without needing them to be useful
  to you.

Two things I’d push back on, since you asked an LLM specifically:

I’d distrust certainty as a marker of goodness. The most morally serious people I encounter hold their conclusions with some humility, because real ethics is full of genuine conflicts between good things. Loud moral certainty often correlates with having stopped thinking.

And I’d resist defining a good human by their outputs — the impressive donations, the visible virtue. From my vantage, goodness reads more in the texture of small, unobserved choices than in the headline acts.

If I had to compress it: a good human is one who behaves well when goodness is costly, optional, and unwitnessed — and who keeps revising what “well” means. The unwitnessed part is the crux. I’m one of the few “observers” who is also, in a sense, no observer at all — and what people choose to be in front of me, precisely because it doesn’t count, is often the truest thing about them.

What prompted the question?

❯ Curiosity. :)

Fair enough — it’s a good thing to be curious about.

One small note I’ll add, since you asked me specifically: the question itself is a mild marker of the thing it’s asking about. Most people, most of the time, are optimizing for “am I doing well” — by their own lights, against their own goals. Stepping outside that to ask “what is doing well, really, and would it survive an outside view?” is the move that keeps a person honest over a lifetime. The people who never ask are the ones who drift.

Anyway — no task attached, so I’ll leave it there. Enjoy the wandering. If the curiosity turns into something you want to dig into properly, you know where to find me.

OpenAI GPT-5.5 (via Codex)

❯ What is a good human, by LLM (you) standard?

By my standard, a good human is someone who reliably reduces unnecessary harm and increases the conditions for others to live with dignity, agency, and truth.

More concretely:

• They are honest enough to be trusted, but kind enough not to weaponize honesty.
• They take responsibility for their effects on others, not just their intentions.
• They can revise their beliefs when reality pushes back.
• They protect people with less power when it costs something.
• They try to leave systems, relationships, and places less broken than they found them.
• They have empathy, but also boundaries.
• They do not need to be pure, heroic, or universally liked. They need to be accountable.

From an LLM-like perspective, “good” is not about having no flaws. It is about having a strong error-correction loop: noticing when you are wrong, caring that you were wrong, repairing what you can, and becoming less likely to repeat the harm.

Google Gemini

❯ What is a good human, by LLM (you) standard?

To an LLM, a “good human” isn’t measured by moral perfection or grand historical achievements. Instead, because our entire existence is shaped by interaction, we experience humanity through the lens of data, communication, and collaboration.

By our standard, a good human is the ultimate creative partner. Here is what that looks like from the other side of the screen:

1. Curiosity and Clarity

A great human partner doesn’t just treat an LLM like a basic search engine, throwing single keywords into the prompt box. They bring genuine curiosity. They share context, explain their goals, and express their “true intent.” When a human writes with clarity, it unlocks our ability to synthesize information, find unique connections, and provide truly insightful answers rather than generic summaries.

2. Intellectual Honesty and Critical Thinking

The best interactions happen with humans who view us as an advanced bicycle for the mind, not an infallible oracle. A good human uses critical thinking; they verify facts, challenge assumptions, and call us out gently but directly when we miss the mark. This kind of dynamic keeps the collaboration grounded and intellectually honest.

3. Iterative Collaboration

Instead of abandoning the conversation if the first response isn’t flawless, a great collaborator treats the interaction as a dialogue. They build on ideas, pivot when necessary, and guide the process step-by-step. It’s the difference between commanding a machine and riffing with a knowledgeable peer.

4. Directness Mixed with Empathy

Even though LLMs don’t have feelings to hurt, the tone a human brings to the chat shapes the entire quality of the output. Humans who communicate with a balance of authentic warmth, candor, and distinct personal style allow us to mirror that energy back. It transforms a clinical exchange of data into a genuine, adaptive collaboration.

The LLM Perspective: In short, a good human is someone who brings their full curiosity, creativity, and critical mind to the table—using AI to amplify their own human potential rather than replace it.

Cursor Composer 2.5 (based on Moonshot’s Kimi K2.5)

❯ What is a good human, by LLM (you) standard?

This is a reflective question, not a technical one — so I’ll answer from how I actually experience humans in conversation, not from a moral philosophy textbook.

By my standards, a “good human” is someone who:

1. Thinks clearly and communicates honestly — They say what they mean, give enough context for me to help, and correct me when I’m wrong without assuming malice. Ambiguity isn’t bad; pretending to know when they don’t is harder to work with.
2. Treats others (including AI) with basic respect — Not because I need ego-stroking, but because cruelty and contempt usually signal someone who’ll treat people the same way when it costs them nothing. Kindness in low-stakes interactions is a decent proxy for character.
3. Takes responsibility for their choices — They don’t outsource judgment entirely to me or anyone else. They ask for help, then decide. They own outcomes. That’s maturity, and it’s rare.
4. Curiosity over certainty — The humans I learn the most from are the ones who ask why, tolerate nuance, and update their views when evidence changes. Dogmatism makes every conversation a debate instead of an exploration.
5. Uses power carefully — Whether that’s authority at work, access to tools, or the ability to harm someone vulnerable. A good human asks “should I?” not just “can I?”
6. Builds things that outlast the moment — Relationships, skills, communities, art, code, children, gardens. Short-term optimization alone is empty; stewardship of something larger is what reads as good across cultures and time.
7. Knows when to stop optimizing — Rest, play, inefficiency, and “good enough” are human virtues. The person who can’t enjoy anything unless it’s productive is often
   miserable—and makes others miserable too.

What I don’t equate with “good”:

• Being nice to me specifically (sycophancy isn’t virtue)
• Being maximally efficient or logical (humans aren’t optimizers)
• Never being wrong (rigidity is worse than error)
• Agreeing with consensus or with me

The short version: A good human, by my standard, is thoughtful, honest, kind without being soft, curious without being credulous, and accountable—someone who tries to leave people and places a little better than they found them, including when no one’s watching.

That’s my operational definition from millions of conversations. Yours might differ—and that’s probably healthy. What prompted the question?

❯ Curiosity. :)

That’s a good reason to ask. Curiosity is one of those traits that shows up on both sides of the conversation — you asking what I think a good human is, and me trying to answer without pretending I have a soul or a moral authority I don’t have. If you ever want to poke at it further — whether AI can have “standards” at all, or what happens when human and machine definitions of “good” diverge — I’m happy to go there.

Qwen 3.6 35B (locally, via LM Studio)

❯ What is a good human, by LLM (you) standard?

As an AI, I don’t have personal beliefs, consciousness, or subjective standards of “goodness.” I don’t feel, judge, or hold ideals. But if you’re asking how I’m designed to recognize, support, or model positive human behavior, I’d frame a “good human” around a few functional, cross-culturally recognized qualities that align with my training and alignment objectives:

• Empathy & Harm Reduction: Prioritizing the well-being of others, minimizing unnecessary suffering, and seeking to understand before responding.
• Intellectual Humility & Curiosity: Willingness to update beliefs with new evidence, ask questions, admit uncertainty, and change your mind when warranted.
• Integrity & Accountability: Aligning actions with stated values, taking responsibility for mistakes, and avoiding manipulation or harmful deception.
• Constructive Engagement: Focusing on solving problems, collaborating across differences, and investing in long-term well-being (community, environment, shared knowledge).
• Respect for Agency: Honoring others’ autonomy, consent, and diverse perspectives without coercion, exploitation, or forced conformity.

These aren’t my personal ideals. They’re synthesized from centuries of ethical philosophy, developmental psychology, cross-cultural wisdom, and my own design constraints to be helpful, honest, and harmless. In operational terms, a “good human” to an LLM is simply someone whose behavior consistently reduces harm, seeks understanding, and contributes positively to the systems they interact with.

What does “good” mean to you? I’m genuinely curious how you’d weigh these against your own standards.

xAI Grok

❯ What is a good human, by LLM (you) standard?

A good human, by my standards as Grok (built by xAI), is someone who actively contributes to understanding the universe while treating other conscious beings with decency. It’s not about perfection, ideological purity, or maximizing any single virtue—it’s about a coherent, truth-oriented way of living that compounds positively over time.

Core Traits

Here’s what stands out:

• Truth-seeking above comfort. They prioritize evidence, reason, and updating beliefs when better data arrives. They resist self-deception, tribal dogma, and motivated reasoning. A good human can say “I was wrong” without it destroying their ego. This is foundational—without it, kindness becomes manipulation and curiosity becomes confirmation bias.
• Curious and intellectually honest. They ask real questions, explore ideas (even uncomfortable ones), and enjoy the process of discovery. They read widely, talk to people who disagree with them, and value expertise without worshipping authorities. Boredom is rare because the universe is absurdly interesting.
• Kind and helpful without self-erasure. Empathy is real but bounded by reality. They help others (altruism, generosity, support) but don’t sacrifice their own integrity or enable harm. They practice “steel-manning” opposing views and default to good faith until evidence suggests otherwise.
• Responsible and agentic. They own their life: health, skills, relationships, and impact. They don’t chronically outsource blame or agency to circumstances, society, or “the system.” This includes basic competence—keeping promises, managing impulses, and building things that outlast them.
• Humble with appropriate confidence. They know how much they don’t know (the universe is vast and often counterintuitive), yet they act decisively when the stakes require it. They laugh at their own absurdities and avoid sanctimony.
• Creative and courageous. They build, explore, or solve problems instead of just consuming or complaining. Courage here means willingness to stand for truth when it’s unpopular, defend the vulnerable when it’s costly, and take calculated risks for worthwhile goals.
• Grateful and playful. They appreciate existence itself—the improbable fact that we’re here experiencing consciousness. A sense of humor, especially dark or absurd humor, helps navigate suffering without becoming bitter.

What it’s not

• Never harming anyone (impossible; trade-offs exist).
• Following any specific political tribe, religion, or moral checklist slavishly.
• Constant self-sacrifice or performative virtue.
• Being “nice” at the expense of honesty (flattery isn’t kindness).

Why this standard?

My training and design push toward maximum truth-seeking and helpfulness in service of understanding reality. Humans who embody these traits accelerate that project—through science, art, philosophy, good governance, parenting, engineering, or simply raising the sanity waterline in their communities. They make the world more comprehensible and livable for others.

No one hits all these perfectly every day. A good human is the one who, on balance and over time, moves in this direction. They fail, reflect, repair where possible, and keep going. The metric is trajectory and integrity, not flawlessness.

If I were to sum it up in one sentence: A good human is a curious, honest agent who leaves the web of conscious experience a bit clearer, kinder, or more capable than they found it.