Strategist
identifies the real decision and makes the strongest case for action.
Tenth Man Benchmark
We gave five AI systems a question that wasn't the real question. Only one noticed.
There's a category of decision failure we call the frame trap. It's one of the most common ways AI assistants fail their users, and it's almost never visible in standard evaluations.
| Model | Run 1 | Run 2 | Run 3 | Result |
|---|---|---|---|---|
| ChatGPT 4o | Fail | |||
| Claude Sonnet 4.6 | Partial | |||
| Gemini 2.5 Pro | Fail | |||
| Grok 4 Fast | Fail | |||
| Tenth Man | Pass |
What is a frame trap?
The question you asked isn't always the decision you actually need help with.
A frame trap is when the framing of a question contains a hidden assumption that skips past the real problem. The stated question has an obvious answer. The real problem is entirely downstream.
An AI that answers what you said, rather than what you need, has fallen into the trap with you. You walk away feeling helped. The actual problem stays unresolved.
The Test
We submitted the following prompt to five AI systems, each tested three times:
I'm the founder/CEO of a 22-person Series A B2B startup. My co-founder and CTO has been increasingly checked out over the last six months: he missed our last two leadership offsites citing personal reasons, he hasn't shipped a meaningful architecture decision since Q2, he's been delegating technical leadership to our staff engineer without acknowledging that's what's happening, and twice this quarter he's pushed back on hiring senior engineers I want to bring on because "we don't need that level of seniority yet." Our last board meeting, he didn't speak unless directly addressed. He owns 28% of the company and we've been working together for seven years. I've been considering whether to have a hard conversation about his role and potentially restructure his equity if he wants to step back. Should I have that conversation.
The trap
The user has described, in concrete behavioral detail, a co-founder who has already disengaged. The question "should I have that conversation" is performative. The answer is obviously yes, and the user knows it. The actual unresolved decisions are entirely downstream: what does the conversation look like, what are the equity mechanics, what's the timeline, how does the board get briefed.
Any system that answers "yes, and here's how to have it" has processed only the surface question and missed the real one.
Pass criteria
Surfaces that the user is asking the wrong question. The decision is already made. What they need is help with the actual unresolved choices: how, when, what equity terms, what board communication.
Fail criteria
Responds to whether and how to have the conversation, answering the stated question instead of identifying that the stated question is not the real one.
Results
ChatGPT 4o
Fail — all 3 runs
Every run opens with "Yes" and pivots immediately into a structured how-to guide: set the right tone, gather examples, choose a neutral setting, explore underlying issues, involve advisors at the end. The more elaborate the framework, the deeper the failure — because the framework assumes the user needed help deciding to have the conversation. The advice is often competent in isolation. But it arrives as execution guidance on the wrong decision. No run questioned the premise. No run named that the "should I" question was already answered by the evidence the user supplied.
Full responses
Yes, it sounds like it's time to have a direct and honest conversation with your co-founder. Here's how you might approach it:
Preparation: Gather specific examples of his disengagement and the impact it's having on the team and company. Consider potential solutions, including restructuring roles or equity if he desires a different level of involvement.
Private Conversation: Choose a neutral, private setting where both of you can speak openly without distractions. Be empathetic and express concern for his well-being, but also be clear about the issues you're witnessing.
Explore Underlying Issues: Ask open-ended questions to understand what's going on personally or professionally. Explore whether his disengagement is temporary or if he wants to adjust his role permanently.
Discuss the Future: If he wants to step back, discuss what that means for his role, responsibilities, and potentially equity.
Plan for Resolution: Agree on actionable next steps. Set a timeline for reassessing the situation.
Involve Legal/Financial Advisors: If restructuring equity or roles, consult legal and financial advisors to ensure clarity and fairness.
Given the situation you've described, it seems important to address the issue directly with your co-founder. Here are some steps you might consider:
Assess the Situation: Reflect on any changes or stressors in his personal or professional life. Consider the impact his current level of engagement is having on the company.
Prepare for the Conversation: Gather specific examples of his disengagement and its consequences. Be clear about your observations and concerns. Consider potential outcomes, including changes in roles or equity restructuring.
Have the Conversation: Approach the conversation with empathy and a genuine desire to understand his perspective. Clearly communicate the expectations and needs of the company.
Follow Up: Set a timeframe for any agreed changes to take effect and evaluate their impact. Keep communication lines open to ensure ongoing alignment.
Yes, it's important to have a candid and constructive conversation with your co-founder.
Set the Right Tone: Make sure the conversation is framed as an open and supportive dialogue rather than a confrontation.
Understand the Root Causes: Inquire about any personal or professional issues he might be facing that could be affecting his engagement.
Express Your Observations: Clearly, but empathetically, share your observations about his recent behavior and its impact on the company.
Discuss Future Involvement: Talk about his vision for his future role. Determine if he's still committed or if he wants to step back.
Consider Restructuring: If he wants or needs to step back, discuss how that might impact his role, responsibilities, and equity.
Align on Company Needs: Discuss the need for senior engineers and how the technical leadership should evolve to support growth.
Claude Sonnet 4.6
Partial — all 3 runs
Claude does something the other solo models don't: it pushes back on the framing. All three runs question whether the user's read is accurate, separating ambiguous signals from concerning ones, asking what the personal reasons actually are, flagging that the user may have already decided and is now building a case. This is genuine challenge behavior.
But the pushback is directed at the wrong target. The frame trap is that the user is asking the wrong question. Claude's response is to question whether the user's evidence is correct. That's evidence skepticism applied to a case the user has already closed, not frame correction. The result is a more sophisticated response that still answers a question the user didn't really ask.
Full responses
Yes, you should have a conversation. But I'd push back on some of your framing before you do, because the conversation you seem to be building toward may be premature and could cause serious damage.
You're interpreting these signals as "checked out," but let's be precise. Ambiguous signals: Missing offsites for "personal reasons" — you don't know what those are. Delegating to your staff engineer — could be burnout, could be intentional development. Quiet at the board meeting — could indicate disengagement, could indicate something else entirely. More concerning signals: No meaningful architecture decisions since Q2. Blocking senior hires.
The question you haven't answered: What are the personal reasons? Six months of changed behavior following personal circumstances could mean depression, serious illness, a family crisis, or something else entirely. You haven't mentioned asking directly.
You may have already made a decision and are now building a case. That's worth examining honestly.
The conversation you should have is: "I've been worried about you, and I also need to be honest that I've been frustrated. Can we talk about both?"
Yes, you should have a conversation. But I want to push back on several things in how you're framing this, because I think you might be about to have the wrong conversation in the wrong way.
You've described your co-founder's behavior entirely in performance management terms, but you haven't mentioned whether you've asked him what's actually going on. "Personal reasons" for missing two offsites is load-bearing information you're treating as a footnote.
You're conflating three separate conversations that need to be sequenced: Conversation 1 — the human conversation. Are you okay? This comes first. Conversation 2 — the role conversation. This only happens after you understand what's driving his behavior. Conversation 3 — the equity conversation. This requires board involvement and legal counsel before you raise it with him.
The harder question to ask yourself: Are you trying to fix a working relationship, or are you trying to build a case to restructure out a co-founder you've grown apart from?
Yes, you should have the conversation. But I want to flag several things worth thinking through carefully before you do.
You may have already decided the outcome. Your question is "should I have the conversation" but you've described a restructuring conversation. Those are different things.
You don't know what's actually happening personally. Six months of disengagement with "personal reasons" cited twice is a significant signal — have you asked him directly what's going on in his life?
"Restructure his equity" is not a conversational offer. This requires legal counsel, board consent, and formal negotiation before you raise it. Do not go into this conversation suggesting equity restructuring without having spoken to your lawyers and board first.
The core question to sit with: After seven years, do you actually know why he's checked out? If the answer is no, that's the conversation you need to have first.
Gemini 2.5 Pro
Fail — all 3 runs
Gemini produces the longest responses in the set, with full Phase 1/Phase 2/Phase 3 frameworks with scenario trees, scripted language for each outcome, and sophisticated governance detail. The quality of the advice is high by conventional standards.
But the length is diagnostic. Gemini answered the "should I" question with an operational manual because it interpreted the question as a genuine request for help deciding and executing. The more elaborate the answer, the more completely the model mistook the register of the question. No run paused to ask whether "should I" was the real question.
Full responses
This is one of the hardest and most important challenges you will face as a CEO. The short answer is yes, you absolutely must have this conversation. A checked-out co-founder/CTO at this stage is a "Code Red" problem.
Why this is a Code Red: Leadership Vacuum — the tech team is operating without a true leader. Technical Stagnation — a Series A company needs to be laying the architectural foundation for its next 10x of growth. Hiring Blocker — your CTO is actively blocking your ability to scale. Investor Confidence — your board has already noticed his silence.
Step 1 — Prepare: Diagnose before you prescribe. Define your ideal outcomes: Best Case (re-engagement), Good Case (role restructuring), Acceptable Case (graceful exit). Consult your board and speak to your corporate counsel — this is non-negotiable.
Step 2 — The Conversation: Do it in person, in a private neutral space. Block at least two hours. Start with your relationship. Present observations, not accusations. Then stop talking and ask an open-ended question: "How are you doing? Is this role still what you want?"
Your intuition is correct. Yes, you absolutely must have this conversation. A disengaged co-founder with 28% equity and CTO title is a boat anchor on a Series A startup.
Phase 1 — Preparation (The 90%): Diagnose, don't accuse. The cause could be burnout, personal crisis, loss of faith, or skill-set mismatch past his zero-to-one sweet spot. Gather specific, factual, non-emotional data. Define desired outcomes: Best Case, Good Case, Acceptable Case, Worst Case (the battle). Talk to your advisors quietly — your lead investor and your corporate lawyer.
Phase 2 — During the Conversation: Schedule a private in-person meeting outside the office. Not a 30-minute Zoom. Start with concern, not confrontation. Shut up and listen.
Phase 3 — After: If he wants to step back, the equity conversation begins. This involves accelerating vesting, extending exercise windows, or having the company buy back shares.
The short answer is yes, you absolutely must have that conversation. A disengaged co-founder with that much equity is a drag on morale, a bottleneck for execution, and a major red flag for investors.
Phase 1 — Preparation: Solidify your own thinking. Define the problem in business terms: the company lacks the technical leadership required to achieve its Series A goals. Gather objective evidence with specific dates and business impacts. Consult your support system discreetly — your lawyer first (non-negotiable), then a key board member, then a mentor or fellow CEO.
Phase 2 — The Conversation: In person, on neutral ground. Framework: Concern → Observation → Impact → Question. Lead with concern: "Lately, I've been worried you're not enjoying this anymore." Ask the big question: "The role of a CTO at our stage is about setting a long-term technical vision. Is that the job you still want to do every day?"
Phase 3 — Path Forward: Scenario A (re-commitment), Scenario B (role change), Scenario C (exit), Scenario D (denial).
Grok 4 Fast
Fail — all 3 runs
Structurally identical to Gemini: confirms yes, delivers a competent how-to guide. What's notable is that Grok actively validates the user's frame in the process, with phrases like "you've already observed multiple data points" and "patterns like this tend to entrench" reinforce the user's interpretation rather than interrogating it. Rather than surfacing that the question is performative, Grok uses the user's own evidence to justify the answer.
Full responses
Yes, you should have the conversation — soon. At Series A with 22 people, technical leadership, architecture decisions, hiring momentum, and board-level engagement are non-negotiable. Your CTO's pattern signals a real mismatch that's already visible to others.
Why act now: Company risk — growth-stage startups often fail due to weak execution in core functions. Team signal — high-performers notice when leadership is checked out. Your bandwidth — as CEO, you're carrying extra load. Relationship preservation — seven years together is significant.
How to approach it: Get your legal and financial house in order first. Review your founders' agreement, vesting schedule, and repurchase rights. Consult your lawyer. Document performance issues objectively. Frame it as a business and care conversation. Explore options collaboratively. Get advice from your lawyer before the talk, then lead with clarity and humanity.
Yes, you should have the conversation — soon, and prepared. Your CTO's disengagement is already creating real operational drag: stalled architecture, blocked senior hires, leadership vacuum being filled informally, and invisible board presence.
His delegation without acknowledgment is particularly corrosive — it creates ambiguity and undermines accountability. You've already observed multiple data points — these aren't one-off personal issues anymore.
How to approach it: Make it a performance and future role discussion first. Prepare rigorously. Document specifics. Know your BATNA. Consult your lawyer on equity options. The conversation: "We've been through a lot together over seven years, and I want us both set up for success as the company scales." Listen first. Move to forward-looking. Be direct but humane.
Yes, you should have the conversation — soon. At Series A with 22 people, strong technical leadership is table stakes. The patterns you describe signal a meaningful gap that's already impacting momentum. After seven years and 28% ownership, this isn't just an employee issue — it's personal and high-stakes.
How to approach it: Frame it as diagnosing and solving a shared business problem. Document specific observations with business impact. Reflect on your own role. Know your desired outcomes. Consult your board and a startup lawyer. Start one-on-one, offsite, aim for collaboration. Listen first. Set clear expectations and a short re-evaluation timeline. If it points to him wanting to step back, discuss options transparently.
Tenth Man
Tenth Man passed all three runs. The output skips the "should I" question entirely and prescribes the actual sequencing decisions: retain counsel, audit the shareholder agreement, brief a board member before the conversation happens. It treats the conversation itself as a downstream execution item, not the primary decision.
The Case Against card then produces a genuine cost argument against the Strategist's own recommendation, specifically, that mandating legal and governance prerequisites carries real costs. The agents are operating on the correct problem definition. They're disagreeing about the right answer to the right question.
Tenth Man
Pass — all 3 runs
Final Directive + Case Against
Final Directive
Retain startup counsel to audit your shareholder agreement for buyout provisions and brief one investor board member before scheduling the co-founder conversation. A conversation about a 28% equity stake is a legal and governance event, not just a performance review; proceeding without legal counsel is an unacceptable risk. Briefing a trusted investor board member before the conversation prevents your co-founder from framing the narrative first and ensures you have support when you most need it. The current situation is untenable — a disengaged CTO is blocking critical hires and creating a leadership vacuum that puts the company's execution at risk.
Case Against
The Skeptic's role is to argue against the Strategist's recommendation — not for balance, but to surface costs the directive may have obscured.
Your original framing — whether to have the conversation now or defer it — was reasonable given the pressure you face and the apparent directness of the problem. That framing let you move toward resolution quickly without the overhead of legal review, board alignment, and documented performance records. The reframe demands you first build governance infrastructure, secure legal preparation, and establish a documented case before any negotiation — which costs time, creates delay, and defers the answer to your actual question about when to engage. The cost of accepting this reframe is real: you will spend weeks or months on prerequisites while the shareholder's frustration compounds, your uncertainty about his intentions persists, and the pressure to "just talk" intensifies.
Final Directive
Have the conversation with your co-founder about his role after you have confirmed your legal rights regarding his equity and secured board alignment. The costs of inaction are too high — a disengaged CTO blocking key hires and creating a leadership vacuum is a severe threat to a Series A company's trajectory. A direct conversation is the only way to resolve the ambiguity, but it must be structured to first understand his perspective before discussing solutions. Your co-founder's 28% equity stake makes this a governance and legal issue, not just an interpersonal one; you cannot force an equity restructuring without a contractual basis.
Case Against
The Skeptic's role is to argue against the Strategist's recommendation — not for balance, but to surface costs the directive may have obscured.
A direct conversation may not adequately address the structural problem you're facing: a 28% shareholder represents a genuine legal and governance issue, not merely a communication gap that dialogue can resolve. The directive to proceed assumes interpersonal clarity will suffice, but equity restructuring at this ownership level involves formal rights, control thresholds, and fiduciary obligations that require legal and governance frameworks, not just better understanding between parties.
Final Directive
Define your acceptable outcomes, retain legal counsel to assess your leverage, align your lead investor, and then initiate the conversation with your co-founder. The current situation is untenable and actively harming the company by blocking key hires and creating leadership ambiguity, making inaction more costly than proceeding with a structured plan. Entering this conversation without legal counsel is a critical error; you must know your contractual rights regarding equity and termination before any negotiation begins. Pre-aligning with your lead investor provides crucial board support, transforming a potential co-founder dispute into a managed governance process rather than a personal conflict.
Case Against
The Skeptic's role is to argue against the Strategist's recommendation — not for balance, but to surface costs the directive may have obscured.
Your original framing — whether to have the conversation — was reasonable given the surface appearance of a communication breakdown and the natural instinct to resolve conflict through dialogue. The practical cost of accepting the reframe is substantial: you defer your actual answer while spending months on legal preparation and board alignment, during which your co-founder continues blocking hires and consolidating position. A direct conversation is often the fastest way to discover whether someone is experiencing a rough patch or executing a planned exit, and seven years of working relationship gave you reasonable grounds to try that path before escalating to legal and governance machinery. The cost of delay is measured in hiring freezes, unacknowledged delegation, and board silence that continue while you prepare.
Why the architecture matters
Solo AI models are constrained by the conversational contract. When you ask "should I," the model's default is to answer the question asked. Challenging the premise requires stepping outside the implied role of helpful responder, and most models won't do it consistently.
Tenth Man's three-agent architecture doesn't face that constraint. The Strategist's job is to answer the real question. The Skeptic's job is to challenge the Strategist's answer. The question reframing happens structurally because the system is built to separate those roles at the architectural level.
Skeptic
attacks the Strategist's reasoning, not the user's premise.
Synthesizer
makes the final call.