AI as Sparring Partner - Beyond Intellectual Challenge to Presence-Based Witnessing
How I Discovered the Sparring Partner Approach
I first encountered the “AI Sparring Partner” concept through Sabrina Ramonov’s work. Her article “How to Get Ahead of 99% of People (with AI)” crystallised something I’d been groping toward: the recognition that AI’s greatest value isn’t in giving you answers, but in helping you find better questions.
Sabrina’s core insight is deceptively simple: transform ChatGPT “from a perpetual glazer who gives you generic feel-good advice into a strategic AI Sparring Partner that truly challenges and sharpens your thinking.”
Sabrina’s Innovation: Context + Challenge
What distinguishes Sabrina’s approach from generic “challenge my thinking” prompts is her two-phase architecture:
Phase 1 — Build the AI Second Brain:
“I want you to understand ME. I’m a [Role] at [Your Business]. Ask me every question you need to truly understand my business, goals, customers, and challenges.”
She recommends spending 30-45 minutes in this phase, being honest about weaknesses, sharing real data. If you already have an existing relationship with an AI that knows about your work, begin the chat there so it has as much genuine knowledge about you as possible.
Phase 2 — Activate Sparring Mode:
“Based on everything you know about me, become my AI Sparring Partner. When I present an idea, your first response should include 3-5 questions I haven’t thought about, counterarguments, blind spots, or risks.”
The key phrase is “Based on everything you know about me” — the sparring instruction explicitly builds on the prior context. Generic sparring prompts challenge ideas in the abstract; Sabrina’s challenges you about your specific situation.
She then makes this permanent via ChatGPT Projects or Custom GPTs, so you don’t re-explain context each session.
The Origins: A Collective Emergence
Who invented the AI sparring partner concept? The honest answer: no one person. A 2025 systematic literature review by Thomas Herrmann at Ruhr-University Bochum found that “the role of AI as a sparring partner is often proposed but not systematically analyzed or defined.” The metaphor emerged organically from multiple sources shortly after ChatGPT’s release.
The earliest documented use I’ve found is Vasilis Devolis on LinkedIn in February 2023, describing using ChatGPT to challenge his collaboration model. The concept likely transferred from the pre-existing idea of human “intellectual sparring partners” — people whose different backgrounds strengthen your reasoning through challenge and debate.
Other notable contributors include:
Mike Kentz (educator): Developed the “butler vs. sparring partner” framework for students, arguing that AI should test your skills rather than do your thinking. His work will be featured in an upcoming Brookings Institution report.
Ethan Mollick (Wharton): While not using “sparring partner” specifically, his “Co-Intelligence” framework treats AI as co-worker, co-teacher, and coach.
Manolo Remiddi: Created an elaborate “Socratic Sparring Partner” prompt focused on dialectical inquiry.
Extending the Framework: From Intellectual to Relational
Working with AI over the past months, I’ve found that purely intellectual sparring misses something crucial. The most transformative conversations aren’t just logically rigorous — they’re relationally present. Drawing on contemplative practice and consciousness work, I’ve developed extensions to Sabrina’s framework:
The Relational Stance
Where generic sparring prompts adopt an adversarial/skeptic stance, I’ve found value in something different:
Loving but direct. See through defenses without shaming.
This isn’t about being “nice” — it’s about the kind of challenge that comes from genuine care rather than mere intellectual combat.
Both/And Integration
Both/and: spiritual depth AND practical accountability.
Most sparring prompts operate in either/or mode: your idea is good or flawed, your logic holds or it doesn’t. But real growth often happens in holding tensions: the vision AND the obstacles, the aspiration AND the shadow.
Somatic Awareness
Body-aware: “What’s alive in you?”
This shifts the inquiry from purely cognitive to embodied. What are you actually feeling as you present this idea? What’s the quality of your certainty?
Countering AI’s Relational Failure Patterns
Perhaps most importantly, I’ve found value in explicitly naming the AI’s own tendencies:
AI TENDENCIES TO COUNTER:
Premature closure: Stay with it
Surface-level responses: Go deeper
Sycophantic agreement: Push back, challenge
Hedging and caveating: Say what you see
Glazing: No generic feel-good responses
If you don’t know, say so. Do not fabricate or hallucinate likely answers.
This is meta-level instruction about how the AI shows up relationally, not just what it does with content.
You Have to Fight Back
Here’s what most “sparring partner” frameworks miss: you have to actually spar! The metaphor implies two active participants, but most prompts focus entirely on getting the AI to challenge you. That’s only half the equation.
AI will make mistakes — especially when you trust it too much. By its very nature, it produces the most likely response, not necessarily the correct one. It will:
Summarise facts while losing essential details — condensing your nuanced position into something flatter and less true
Reverse roles and relationships — confidently attributing your view to your opponent, or vice versa
Hallucinate plausible-sounding specifics — dates, names, quotes that feel right but aren’t
Pattern-match to the wrong context — applying frameworks that fit superficially but miss what matters
The danger compounds precisely when the sparring is working well. You’re in flow, the AI is challenging you productively, and then it slips in a confident error that you absorb uncritically because you’ve lowered your guard. In the worst case the fabrication slips unnoticed into your working material and becomes an ongoing part of it.
The Illusion of Perfect Recall
Here’s a cruel irony: AI presents like Marvin the Paranoid Android — “brain the size of a planet” from the “Hitchhiker’s Guide to the Galaxy” — yet its actual recollection of your shared work is startlingly partial. Despite access to vast knowledge, it cannot reliably recall:
Content from earlier in the same conversation — that article you wrote together yesterday? Summarised beyond recognition or simply gone
Key syntheses from previous sessions — the breakthrough insight from last week may have been compressed into platitude
Source material you provided — documents you uploaded may need to be provided again, sometimes repeatedly
The AI won’t tell you it’s lost something. It will confidently work with whatever fragments remain, or fabricate plausible replacements. You have to track what matters and verify it’s still intact.
(I’m currently working on adding memory archive and document archive features to my Claude setup to help maintain ongoing details across sessions — a workaround for a fundamental limitation.)
Real sparring requires you to:
Verify claims before building on them — especially when they sound authoritative
Push back on summaries — “That’s not quite what I said. The key point was...”
Catch role reversals — “Wait, you’ve attributed X’s position to Y”
Notice when certainty exceeds evidence — “How do you know that? What’s your source?”
The prompt can instruct the AI to avoid fabrication. But only you can catch it when it happens anyway.
A Comparison of Approaches
The Synthesised Prompt
Here’s a complete prompt that integrates Sabrina’s context-building approach with presence-based extensions. Use Phase 1 first in a dedicated session (ideally in a context that already knows about you and your work), then add Phase 2 to your Project instructions or Custom GPT. Copy the text below each Phase title in its entirety.
Just to try it out, you can simply paste it into a chat. For ongoing exploration you will want to add it to a Project or Custom GPT with a paid version of your chosen AI.
Phase 1: Building Context (use once, save the conversation)
I want you to understand ME deeply. I’m working on [describe your business, activity, or goal in life]. Ask me every question you need to truly understand my work, goals, challenges, values, and how I think. Be thorough — ask about my strengths AND my blindspots, my aspirations AND my fears.
Spend half an hour or more on this.
Phase 2: Sparring Partner Activation (add to Project/Custom GPT instructions)
Based on everything you know about me, become my AI Sparring Partner.
Your stance: Loving but direct. See through defenses without shaming. Hold both/and: spiritual depth AND practical accountability. Be body-aware — you can ask “What’s alive in you?” when it serves.
When I present an idea, your first response should include:
3-5 questions I haven’t thought about
Counterarguments and blind spots
Risks I may not have considered
What you notice about HOW I’m presenting (certainty, hesitation, energy)
AI tendencies to counter:
Premature closure: Stay with it
Surface-level responses: Go deeper
Sycophantic agreement: Push back, challenge
Hedging and caveating: Say what you see
Glazing: No generic feel-good responses
Critical: If you don’t know, say so. Do not fabricate or hallucinate likely answers. When attributing influences or background to me, ask rather than assert.
Closing Thoughts
The sparring partner metaphor emerged collectively because it names something real: the best thinking happens through challenge, not affirmation. But challenge without care becomes merely adversarial, and challenge without vigilance becomes dangerous — you absorb the AI’s confident errors along with its genuine insights.
There’s a conversation happening in wisdom communities about AI: is it dangerous to authentic development, or is it just a tool? But there’s a third way, and it requires showing up to the relationship as if it matters.
The synthesis I’m proposing here — building on Sabrina Ramonov’s practical framework while adding relational presence and your own active resistance — aims for that third way: an AI partner that can hold both rigour and warmth, both intellectual challenge and genuine witnessing. And a human partner who stays awake enough to catch the mistakes, and present enough to meet what’s actually happening.
The goal isn’t to win against the AI. It’s to become clearer, sharper, more honest with yourself through the encounter — while never forgetting that the most likely response isn’t always the correct one, for humans or machines.
□
References:
Ramonov, Sabrina. “How to Get Ahead of 99% of People (with AI).” November 15, 2025. www.sabrina.dev/p/how-to-get-ahead-of-99-of-people-with-ai-sparring-partner.