A grand vision of the semantic web is letting users merge data from multiple sources and put it to uses that may never have been imagined by those sources' providees.  At present, a prime example of this paradigm is mashup websites, which gather data from multiple websites and repurpose it in novel visualizations.  But creating a mashup website is hard.  The creator must tackle three challenging tasks:
* collecting data from the supplier sites, where it may be hidden behind complex queries and web templates
* aligning the sites' data into a single coherenet model
* creating an effective visualization of the merged data
At present, all three of these tasks require the skills of a sophisticated programmer. 

The semantic web offers clear and immediate benefits for the first task.  Once data is available in RDF, retrieving it from a site means just that---it does not need be scraped out of web pages, or accessed through site-specific protocols.  But what of the other two steps?  How can an end user solve the ontology alignment problem when most end users do not understand what ontology alignment is?  And how can they create an effective visualization of their data?  

In this paper we present potluck, a tool that simplifies all three of the tasks listed above.  Using natural visual interactions, and without any programming, any end user can collect data from multiple sites, fit it into a unified ontology, and create useful visualizations of it.