Prompting Best Practices
AI Features /Credits System /Generate a Board with AI /Prompting Best PracticesWrite clearer prompts for denser, more useful boards without wasting effort or credits.
Strong prompts do not need to be long. They need to be clear about the outcome you want, the material Corkulous should use, and the level of detail that actually matters. A clear prompt usually produces a better board than simply asking for “more detail” or “a bigger output.”
If you are not sure what to ask for yet, start with the Prompt Architect Guide. It shows how to use a short planning conversation to turn a rough idea into a Corkulous-ready prompt.
Start with the Job to Be Done
Begin with what the board is for:
- planning a launch
- studying a topic
- summarizing a document
- comparing options
- running a meeting
- building a mood board
That gives Corkulous a target, not just a topic.
Instead of: “Make a board about coffee”
Try: “Create a board to compare home coffee brewing methods for a beginner choosing their first setup.”
Ask for Structure, Not Just More Words
If you want a dense board, say how it should be organized. Useful structure cues include:
- number of sections or clusters
- timeline, workflow, comparison, or hierarchy
- short notes vs fuller cards
- checklists, risks, milestones, or decision criteria
Instead of: “Make it very detailed”
Try: “Create 6 sections: goals, timeline, team roles, budget risks, launch checklist, and success metrics. Keep each card concise but specific.”
Include What Matters Most
Tell Corkulous what must appear on the board:
- required topics
- names, dates, or constraints
- audience
- tone or style
- what to leave out
Examples:
- “Focus on actions for the next 30 days, not long-term strategy.”
- “Use plain language for parents, not technical school jargon.”
- “Do not add images or decorative items.”
Use Attachments Intentionally
When your board-generation workflow includes PDFs or text documents, explain how Corkulous should use them. Good attachment prompts usually say:
- whether to summarize, extract, compare, or reorganize
- which sections matter most
- whether to preserve the source language or rewrite it
- what kind of board should come out the other side
Examples:
- “Use the attached PDF to create a board with executive summary, key charts, risks, and next actions.”
- “Read the attached meeting notes and build a board with decisions, open questions, owners, and deadlines.”
If a document is very large, focus the prompt on the specific parts you care about most.
Use Web-Enhanced AI Only When Current Facts Matter
Turn on Web-Enhanced AI when you need current information such as:
- recent news
- live market conditions
- current pricing
- recent research or product changes
When you do, be specific about what should be researched:
- “Find 2026 pricing and feature differences for the top three email platforms.”
- “Use recent sources from the past 12 months.”
That keeps the board focused and helps avoid wasted searches.
Dense Boards Without Runaway Cost
More output is not automatically a better board. A clear, structured prompt usually beats an oversized request. To get a rich board without unnecessary sprawl:
- ask for a specific number of sections
- ask for concise but complete cards
- name the highest-priority topics
- avoid “cover everything” unless you truly need it
- use attachments instead of pasting huge blocks of source material
If the first board is close but too thin, follow up with a narrower refinement prompt instead of restarting from scratch with a much broader request.
Prompt Patterns That Work Well
Planning Board
Create a launch board for a small paid iPhone app update. Include release goals, timeline, QA checklist, App Store assets, announcement plan, and post-launch metrics. Keep it practical and team-oriented.
Research Board
Build a board comparing three note-taking apps for a design team. Include strengths, weaknesses, collaboration features, offline support, pricing, and recommendation criteria.
Attachment-Based Board
Use the attached report to create a board with main findings, surprising data points, risks, and recommended actions. Prioritize the executive summary and conclusion.
Current-Events Board
Create a board on the current state of residential solar incentives in the United States. Use current sources, group by federal vs state incentives, and include a short section on what is changing this year.
Prompt Progression: From Simple to Rich
It helps to see how a prompt can grow without turning into a wall of filler. The best longer prompts add useful constraints, not extra noise.
1. Small Prompt
Make a board about OpenAI's reported Spud model.
This can work, but it leaves too much to chance.
2. Better Prompt
Build a Corkulous board about the reported OpenAI "Spud" model. Compare what is confirmed, what is reported, and what is still unknown. Include OpenAI, Anthropic Claude Opus 4.7, likely use cases, and enterprise implications.
This gives the board a clearer job and a better comparison target.
3. Strong Research Prompt
Build a dense Corkulous board about the reported OpenAI "Spud" model and the current enterprise AI race around it. Treat it as a current-events research board, so use web search and pull from recent sources. Compare Spud with Anthropic Claude Opus 4.7. Include what is confirmed, what is reported, what is speculative, and what is still unknown. Add sections for strategy, model capabilities, enterprise positioning, risks, and likely release scenarios.
Now the board knows it should research, compare, and separate evidence from rumor.
4. Rich, High-Constraint Prompt
Use a longer prompt when the board really needs a strong point of view, tighter layout guidance, and richer source handling:
Build a dense, visually rich Corkulous board about the reported OpenAI "Spud" model and the current enterprise AI race around it. Treat this as a current-events research board, so you must use web search and pull from recent sources, including official vendor pages, major reporting, and technical docs. Compare the reported Spud model with Anthropic Claude Opus 4.7, and include what is confirmed, what is reported, what is speculative, and what is still unknown. Make the board feel like an analyst's wall: evidence map, timeline, product strategy, model capabilities, enterprise positioning, safety questions, likely release scenarios, and a practical testing plan for evaluating Spud the moment it ships.
Make the board dense, creative, and pretty, not sparse. Use a lot of item variety: labels, cards, notes, shapes, tasks, link previews, and at least one meaningful sub-board if there is enough content to justify it. Do not bury sources in plain text notes; use real link preview items for key sources and distribute them naturally through the layout. Include sections for OpenAI strategy, Anthropic's current position with Opus 4.7, comparison tables, watch items, rumors versus official signals, and a launch-day response playbook. Add a flowchart showing how to evaluate whether Spud is worth adopting, routing to, or ignoring. Include specific facts, dates, pricing if available, rollout clues, benchmark claims only when sourced, and a final verdict that clearly separates evidence from hype.
Use recent sources only where it matters, and prioritize authoritative ones. Pull from OpenAI official pages, Anthropic official pages, and major reporting outlets covering the Spud story and Opus 4.7. Make the board comprehensive enough that it feels like a serious frontier-model briefing rather than a light summary. Keep the layout tight, layered, and information-dense, with strong visual hierarchy and minimal wasted space.
This works because it does four useful things at once:
- defines the board’s job
- tells Corkulous where to get information
- describes the visual and structural feel
- separates evidence from speculation
Another Rich Example
This same pattern also works well for policy, markets, and infrastructure topics:
Build a dense, visually rich Corkulous board about the current AI chip war, centered on U.S. export controls on advanced chips to China and how NVIDIA, AMD, TSMC, Microsoft, OpenAI, and major Chinese players are responding. Treat it as a current-events research board, so you must use web search and rely on recent reporting, official company statements, earnings materials, policy documents, and technical coverage. Make the board feel like an analyst war room: policy timeline, company strategies, supply-chain chokepoints, product comparisons, geopolitical pressure points, likely winners and losers, and what this means for AI model development over the next 12 months.
Make the board dense, creative, and pretty, with strong visual hierarchy and very little wasted space. Use a wide mix of items: labels, cards, notes, shapes, tasks, link previews, comparison tables, and at least one meaningful sub-board if there is enough content. Do not bury sources inside plain text notes when link previews would work better. Include sections for export-control policy changes, NVIDIA and AMD product workarounds, TSMC manufacturing dependence, hyperscaler demand, China's domestic chip push, investor and market reaction, and practical implications for AI labs and app developers. Add a flowchart showing how a company might decide between using frontier chips, downgraded export-compliant chips, cloud rentals, or alternate suppliers. Separate confirmed facts, reported developments, and speculation clearly, and end with a practical verdict about how the chip war could reshape the frontier-model race.
This version works for the same reason: it gives Corkulous a research task, a board structure, a visual feel, and a clear standard for what counts as evidence.
When you write longer prompts, try to make every sentence earn its place in the board.
When the First Result Is Close
The best follow-up prompts are surgical. Try requests like:
- “Keep the same structure, but make the timeline section more concrete.”
- “Reduce the number of sections and combine overlapping ideas.”
- “Add more decision criteria and tradeoffs.”
- “Replace generic notes with actionable next steps.”
- “Keep it dense, but shorten every card to one compact paragraph.”
That usually works better than starting over with a vague “try again.”
