Most AI interactions are transactional. You have a problem, you ask, you get an answer, you close the tab. That works fine. But there’s a different mode — using it as a recurring sounding board rather than a one-off search engine — that almost nobody sets up deliberately, and it’s the one that actually compounds over time.
What a Sounding Board Actually Is (vs. What Most People Do)
Most people use AI the way they’d use Google with extra steps: get the output, move on. A sounding board is something else. It’s a place where you think out loud, repeatedly, about something you’re working through over days or weeks — and where the friction of articulating your thinking to something that pushes back actually changes the thinking.
The distinction isn’t about the tool. It’s about how you’re using it. A single session asking Claude to review your business idea is a one-shot query. Returning to it three times over two weeks, each time with new information or a shifted position, and each time asking it to challenge whatever you currently believe — that’s a sounding board.
The reason this improves with repetition is simple: you get better at describing the problem. The first time you explain something, you’re usually explaining the surface version. By the third time, you’ve noticed the parts that don’t hold up, you’ve updated the framing, and your prompt is much closer to the actual question. The model’s response quality goes up not because the model improved — because your input did.
How to Actually Set This Up
The practical obstacle is that AI has no memory across sessions by default. ChatGPT’s memory feature helps, but it’s patchy — it remembers facts you’ve told it, not the texture of ongoing reasoning. Claude doesn’t persist memory at all unless you’re using a third-party integration or doing it manually.
The workaround that actually works: keep a short, living document — a few bullet points — that you paste at the start of any new session on the topic. Not a full transcript. Just: what you’re working through, where you currently stand, and what you want challenged today.
It sounds like overhead. It isn’t, once you have it. Writing the summary forces you to articulate your current thinking, which is itself useful. And the model isn’t starting from zero every time.
A rough template:
- The situation: One or two sentences on what you’re deciding or working through.
- Where I currently stand: Your actual current position, not the question you’re uncertain about.
- What I want from this session: Push back on X, help me think through Y, stress-test the assumption that Z.
That last part is the one people skip. Asking for general “thoughts” gets general thoughts. Asking it to specifically argue against your current position — or find the weakest link in your reasoning — gets something more useful.
Where It’s Actually Useful
Not everywhere. One-shot answers work fine for factual lookups, code snippets, quick edits. The sounding board pattern earns its keep on problems that don’t resolve in one sitting:
- Deciding whether to pursue something — a job change, a project, a freelance client you’re unsure about
- Working through a disagreement with someone where you want to pressure-test your own read before acting on it
- Building a position on something you don’t fully understand yet, where you need to keep updating as you learn more
- Any creative or strategic project where you’re iterating, not executing
The common thread is that your position is supposed to change. If you already know what you think and just want validation, a sounding board won’t help — it’ll agree with you or produce mild pushback you’ll ignore. The value kicks in when you’re genuinely uncertain and want resistance, not reassurance.
The Actual Risk
There’s one failure mode worth naming directly: you can start using the model as a substitute for sitting with discomfort rather than a tool for resolving it. If you’re returning to the same conversation every day and leaving each session feeling temporarily clearer but not actually further along, the sounding board has become a procrastination mechanism.
A real sounding board session should end with something: a clearer position, a specific next action, a question you’ve narrowed down. If you finish a session feeling like you explored the topic but can’t say what changed, the session did something other than help you think.
The model will cheerfully keep engaging regardless. It doesn’t know when you’re going in circles. That’s your job to notice.
One Concrete Thing Worth Trying
Pick something you’ve been sitting on — a decision, a project, a problem at work — that you haven’t been able to fully resolve. Spend five minutes writing down where you actually stand on it right now. Not the history of it, not all the considerations — just your current position.
Then open a new chat and ask the model to argue against that position as specifically as possible. Not “what are the counterarguments” — “here’s what I think, tell me why I’m wrong.”
The first response will probably surface one thing you hadn’t considered, or put pressure on an assumption you were holding loosely. Write down what shifted. Come back tomorrow with the updated version. After three or four rounds, you’ll have a much sharper sense of what you actually believe and why — and whether it holds up under pressure. At that point the question worth asking is whether you’ve moved toward a decision, or just gotten more articulate about your uncertainty.
