Estimating with Claude: Why Your Current Approach Isn't Working
If you're searching for a Claude prompt for estimation, chances are you're trying to use AI to simplify complex calculations or predict uncertain outcomes. Perhaps you're a project manager trying to estimate the time and resources required for a new project, or a business owner trying to forecast sales and revenue. Whatever your use case, you're likely finding that your current approach to estimation is time-consuming, prone to errors, and often inaccurate. This is because estimation requires a combination of data analysis, domain expertise, and critical thinking – skills that are difficult to replicate with traditional prompting methods.
The Pattern that Works
To overcome these challenges, we've found that stacking the L99 and /deepthink prompt codes can be highly effective. The L99 code provides a framework for breaking down complex problems into smaller, more manageable components, while /deepthink encourages Claude to think critically and consider multiple perspectives. By combining these codes, you can create a prompt that not only estimates outcomes but also provides a clear explanation of the underlying assumptions and calculations.
Before and After Example
Here's an example of how this pattern can be applied in practice:
Before: What are the estimated costs of developing a new mobile app?
After: Using L99 and /deepthink, estimate the costs of developing a new mobile app, considering factors such as development time, team size, and technology stack. Provide a detailed breakdown of the costs and explain any assumptions made in the estimation process.
In response to the After prompt, Claude might provide a detailed estimate of the costs, including:
- Development time: 12 weeks
- Team size: 4 developers, 1 designer, 1 project manager
- Technology stack: React Native, Node.js, MongoDB
- Estimated costs: $200,000 - $250,000
- Assumptions: The estimate assumes a moderate complexity app with a standard set of features, and that the development team has prior experience with the technology stack.
Anti-Patterns that Don't Work
While the L99 and /deepthink combination is effective for estimation, there are several anti-patterns that people try which don't work. For example:
- Using the /trim code alone, which can oversimplify the estimation process and lead to inaccurate results.
- Stacking /hook and /simplify, which can create a prompt that is too focused on a specific aspect of the problem and neglects other important factors.
- Applying the SENTINEL code, which is better suited for anomaly detection and monitoring rather than estimation.
When NOT to Use this Approach
While the L99 and /deepthink combination is a powerful tool for estimation, there are certain situations where it may not be the best approach. For example:
- When dealing with extremely complex or nuanced problems that require human expertise and judgment.
- When the estimation requires a high degree of precision or accuracy, such as in financial modeling or scientific research.
- When the underlying data is incomplete, inaccurate, or biased, which can lead to flawed estimates and incorrect assumptions.
Next Steps
To learn more about the L99 and /deepthink codes, as well as other prompt codes that can be used for estimation and other tasks, see all 120 codes tested over 3 months in the Cheat Sheet. This comprehensive resource provides a detailed guide to using Claude for a wide range of applications, from estimation and forecasting to text generation and conversation.