Best Practices for Solve | Rocket.new - AI Research Beyond ChatGPT

Solve produces reports that match the quality of your question. Specific, well-scoped prompts return focused analysis you can act on. Vague prompts return broad, generic summaries. This page covers question framing, follow-up patterns, and iteration strategies that get the most out of every Solve task.

Writing good questions

Think of Solve as a research analyst. Give it context, name constraints, and define your expected output. On Full Solve, Rocket may use prompt intelligence before research begins to resolve ambiguity. Light Solve may ask conversational follow-ups in chat for vague prompts, but it does not use prompt intelligence. You can also reference previous work through @-mentions or attach files and URLs for additional context.

Weak prompts vs. strong prompts

Weak prompts
Strong prompts

Prompt	Why it falls short
”Tell me about the SaaS market”	No segment, geography, or specific question
”Who are our competitors?”	Solve has no context about who “we” refers to or what market you are in
”What should we build?”	No context about your product, users, or constraints
”Is this a good business idea?”	No specifics about the idea, market, or success criteria
”Help with pricing”	No product details, target market, or competitor benchmarks

Prompt	Why it works
”What is the TAM for AI-powered code review tools targeting mid-market engineering teams in North America? Include growth projections through 2027.”	Specific market, segment, geography, and time frame
”Compare Linear, Jira, and Shortcut on features, pricing, target audience, and developer experience. Focus on teams of 10 to 50 engineers.”	Named competitors, defined dimensions, target persona
”We have a B2B SaaS product at $5M ARR with 80% YoY growth. What revenue multiples should we expect at Series B? Pull comparable recent rounds.”	Specific metrics, clear ask, defined comparable set
”Should we price at $12,$ 18, or $25 per user per month for an AI writing tool targeting content marketing teams? Use Jasper and Copy.ai as benchmarks.”	Specific price points, named competitors, defined audience
”Prioritize these 5 features for a CRM targeting real estate agents using a RICE framework: automated follow-ups, MLS integration, AI lead scoring, mobile app, team dashboards.”	Listed features, named framework, specific vertical

Five ways to sharpen your prompt

Include context about your situation

Tell Solve who you are and what stage you are at. “We are a seed-stage startup with 3 engineers targeting SMBs” produces very different recommendations than the same question without that context.Useful things to include:

Your company stage (pre-launch, seed, Series A, and so on)
Your target customer (segment, size, industry)
Relevant constraints (budget, timeline, team size)
What you have already tried or decided

Name names

Instead of “analyze the competition,” name the 3 to 5 competitors you care about. Instead of “research the market,” name the specific category. Specificity turns a broad survey into a focused analysis.Instead of: “What are the trends in AI?”Try: “What are the top 3 trends in AI-powered B2B sales tools for 2025? Focus on products like Gong, Outreach, and Salesloft.”

State the purpose of the answer

A market size estimate for a pitch deck needs different precision than one for internal planning. Tell Solve what the output is for and it will calibrate accordingly.Instead of: “What is the market size for project management tools?”Try: “I need a TAM, SAM, and SOM estimate for project management tools targeting remote-first companies, formatted for a Series A pitch deck.”

Request specific formats and frameworks

If you want a SWOT analysis, say so. If you want a comparison table, ask for one. Explicit format requests prevent Solve from guessing your intent.Useful format requests:

“Create a feature comparison table”
“Use a RICE scoring framework”
“Structure this as a two-page investor memo”
“Give me a pros and cons list for each option”
“Present the bull case and bear case separately”

Set boundaries on scope

Without boundaries, Solve may go too broad or too narrow. State what to include and what to skip.Instead of: “Analyze Stripe”Try: “Analyze Stripe’s pricing model and payment processing fees. Do not cover their banking or identity products.”

Follow-up patterns

Use follow-up messages to extract more value from the initial report. Three patterns work well: drill-down, challenge, and pivot.

Pattern	When to use	Example sequence
Drill-down	You want more detail on one section of the report	1. “Map the competitive landscape for no-code database tools.” 2. “Tell me more about Airtable’s enterprise strategy. What features do they gate behind enterprise pricing, and how does that compare to Notion?” 3. “If I were building a no-code database for agencies, what 3 features would differentiate me from Airtable?”
Challenge	You want to stress-test the conclusions	1. “Build an investment thesis for vertical SaaS in healthcare.” 2. “What is the strongest bear case against this thesis? What would make this investment fail?” 3. “Given those risks, how would you modify the thesis to account for regulatory headwinds?”
Pivot	An unexpected finding redirects your research	1. “What is the market size for AI tutoring tools for K-12 students?” 2. The report reveals the fastest-growing segment is corporate training. 3. “Pivot to the corporate training market. What does the AI-powered training market look like for companies with 500 or more employees?”

Each pattern follows the same rhythm: start broad, react to findings, then narrow or redirect. You can refine any report directly from the results page.

Iteration strategies

Start broad, go narrow

Begin with a landscape question such as “What does the market look like?” then narrow to specifics. This gives you context before committing to a direction.

Research, then decide

Ask factual questions first (“What pricing models do competitors use?”) then ask decision questions (“Which model fits my constraints?”). Separating research from decisions produces better answers for both.

Analyze, then synthesize

Run separate Solve tasks on different aspects of a problem: one for market analysis, one for competitive research, one for pricing. Then ask a final task to synthesize: “Combine these findings into a go-to-market strategy.” Tasks in the same project can reference each other with @-mentions.

Combining Solve with other capabilities

Solve works well alongside Build and Intelligence.

Combination	How it works
Solve then Build	Research a market or feature set with Solve, then use those findings to scope a Build task. For example: “Build a pricing page based on the competitive analysis from my Solve task.”
Solve then Intelligence	Use Solve to identify key competitors, then set up an Intelligence task to monitor them continuously.
Intelligence then Solve	When Intelligence surfaces a significant change, such as a competitor launching a new feature, create a Solve task to analyze the implications.
Solve then Solve	Chain multiple tasks in a project: market analysis first, then competitive teardown, then pricing strategy. Each task can build on previous findings.

All tasks within a project share the same context. You can @-mention a previous task to bring its findings into any new Solve task without re-entering data.

Common mistakes

Mistake	What to do instead
Asking multiple unrelated questions in one prompt	Split into separate Solve tasks, each focused on one topic
Not providing enough context	Include your stage, audience, constraints, and what you already know
Taking the first result as final	Use follow-ups to challenge, refine, and deepen the analysis
Ignoring the methodology section	Read how Solve arrived at its conclusions so you can evaluate confidence
Using Solve when Intelligence is better	If you need ongoing monitoring, set up an Intelligence task instead of re-running Solve

SOLVE

RESEARCH

WORKING WITH RESULTS

Best practices

Writing good questions

Weak prompts vs. strong prompts

Five ways to sharpen your prompt

Follow-up patterns

Iteration strategies

Combining Solve with other capabilities

Common mistakes

What’s next

Quick start

Use cases

​Writing good questions

​Weak prompts vs. strong prompts

​Five ways to sharpen your prompt

​Follow-up patterns

​Iteration strategies

​Combining Solve with other capabilities

​Common mistakes

​What’s next

Quick start

Use cases

Writing good questions

Weak prompts vs. strong prompts

Five ways to sharpen your prompt

Follow-up patterns

Iteration strategies

Combining Solve with other capabilities

Common mistakes

What’s next