Five Habits That Make Any AI Chat Tool Far More Useful
— 2 min read
People often blame the model when an AI assistant gives a mediocre answer. More often the gap is in how the tool is being used. These five habits cost nothing to adopt and compound over time.
1. Lead with context, not the question
"How should I structure this?" gets a generic answer. "I'm a solo founder building a booking app, I use TypeScript, and I want something I can maintain alone. How should I structure this?" gets a useful one. Two sentences of background change the quality of the response.
2. Ask for a format you can use
Say what you need back: a checklist, a table, three options with trade-offs, a draft under 150 words. Specific formats save editing time.
3. Save what worked
When a conversation produces a good template, decision or explanation, do not leave it in chat history. Copy it into a notes file, or into a tool built for the purpose. Coconut is one such tool: it gives your AI assistants a shared long-term memory over MCP, with a separate private notebook for each project, client or person, so a note saved from one assistant can be found from another. A plain document works too. What matters is that the good stuff outlives the chat.
4. Verify anything that matters
Assistants can sound confident while being wrong. For facts, figures, legal or medical points, check a primary source. Ask the assistant to show its reasoning or cite where a claim comes from, then confirm the source yourself.
5. Watch the cost of context
If you pay per use, long inputs add up. Developers feel this most: coding agents can re-read large amounts of code to answer a single question. Approaches such as CodeMesh, which builds a structural map of a repository so the agent can fetch only the relevant parts, aim at exactly that. Even if you never write code, the principle holds: send the assistant what it needs, not everything you have.
Putting the habits together
| Habit | What it fixes | Effort |
|---|---|---|
| Lead with context | Generic answers | Low |
| Request a format | Time spent reformatting | Low |
| Save what worked | Re-explaining every session | Medium |
| Verify important claims | Confident mistakes | Medium |
| Watch context cost | Surprise bills, slow replies | Medium |
Frequently asked questions
Which habit should I start with?
Start with context. It is free, immediate, and improves every other interaction.
How much background is too much?
If most of it is irrelevant to the question, trim it. Relevant and recent beats long.
Do these habits work across different assistants?
Yes. None of them depend on a specific product, which is why a shared memory that works across assistants is worth considering if you switch between tools.