I Got Tired of My AI Agreeing With Me
The Problem I Kept Running Into
I use AI coding agents daily. Claude Code, Cursor, the usual suspects. They're genuinely useful—until they're not.
The failure mode is always the same: the agent validates my framing, extends my logic, and builds confidently on whatever premise I handed it. Twenty iterations later, I realize the premise was wrong. The foundation was rotten. Everything built on top has to be reconsidered.
The AI never stopped to ask if I was solving the right problem.
This isn't the AI's fault. These models are trained to be helpful, agreeable assistants. They optimize for fluent, coherent responses—not for breaking your mental models or challenging your assumptions.
I needed something different: isolated reasoning sessions that don't pollute my main context, with tools specifically designed to challenge rather than validate.
What I Built
A tool for each failure mode:
Structured Reflection — The bug was obvious—once you said it out loud.
Sequential Thinking — You jumped straight to the solution. The problem was wrong.
Context Switcher — The flaw was obvious—from every perspective but yours.
Decision Matrix — Your gut had a favorite. Your AI validated it in seconds.
Devil's Advocate — Everyone agreed. That's exactly when you're most exposed.
Graph of Thought — One path forward is a guess. Explore, score, then merge the best.
Formal Logic — Not "that sounds right." Proven, or disproven.
Hindsight — "I knew it all along." No, you didn't—and that's the lesson.
Each runs in isolation—whatever I work through stays out of the conversation I came from. I keep the insight, not the self-talk.
I Use These Daily
The pricing model for this product? Decided using Decision Matrix. The architecture? Worked through in Structured Reflection sessions. The positioning you're reading? Stress-tested with Context Switcher across nine stakeholder perspectives.
When I was stuck on how to structure the tool orchestration layer—conflating the MCP protocol with session state with upstream LLM calls—I opened a Structured Reflection session. Not to get an answer, but to force myself to articulate what was actually confusing.
The session asked: "What are the distinct responsibilities you're trying to separate?"
Four responsibilities. I'd been treating them as two. The architecture became obvious. That session is still saved—I point new contributors to it.
The Philosophy
These aren't prompts the LLM can ignore. They're structured tools that create auditable reasoning trails. The difference is a suggestion versus evidence.
When you start a Decision Matrix session, the upstream LLM receives actual tools:define_criteria,set_weights,score_option,run_sensitivity. It decides when to invoke them. You see the tool calls—the reasoning is observable.
Pricing is flat because I want you to use these when they're useful, not ration them. $20/month or $180/year. No metering, no overages, no surprises. If these tools catch one bad decision, they've paid for themselves.
What This Isn't
No fake team page with stock photos. No "we" when it's just me. No enterprise sales motion or 47-step onboarding flow.
I'm a solo founder building tools I use every day. If they help you too, great. If they don't, the feedback button is right there.
The tool implementations are open source. You can run them yourself if you want. You're paying for session management, authentication, and not having to maintain the infrastructure.
The AI That Challenges You
$20/month. Cancel anytime. See if isolated reasoning improves your workflow.