As cmhughes suggested in his comment, you can use an enumerate environment with labelled \items to build the theorem, and then use \ref in the proof (this guarantees you consistency and avoids possible errors from manually numbering):
\documentclass{article}
\usepackage{amsthm}
\usepackage{amsmath}
\newtheorem{theorem}{Theorem}
\begin{document}
\begin{theorem}
For a left artin ring $\Lambda$ we have the following.
\begin{enumerate}
\item\label{th:radnil} The radical of $\Lambda$ is nilpotent.
\item\label{th:finsim} There is only a finite number of nonisomorphic simple $\Lambda$-modules.
\item\label{th:leftnoe} $\Lambda$ is left noetherian.
\end{enumerate}
\end{theorem}
\begin{proof}
\ref{th:leftnoe}. Since $\Lambda$ has a finite filtration...\par
\ref{th:radnil}. Let $I$ be an ideal in $\Lambda$...\par
\ref{th:finsim}. If for a $\Lambda$-module $A$ we have...
\end{proof}
\end{document}

Using the enumitem package one can define a customized list-like environment; in the following example the thmclaim environment works as an enumerate environment but using boldfaced labels; inside the proof environment, a description environment was used to refer to the claims (thus keeping consistent use of boldfaced type):
\documentclass{article}
\usepackage{amsthm}
\usepackage{amsmath}
\usepackage{enumitem}
\newtheorem{theorem}{Theorem}
\newlist{thmclaim}{enumerate}{1}
\setlist[thmclaim,1]{label=\normalfont\textbf{\arabic*.}}
\begin{document}
\begin{theorem}
For a left artin ring $\Lambda$ we have the following.
\begin{thmclaim}
\item\label{th:radnil} The radical of $\Lambda$ is nilpotent.
\item\label{th:finsim} There is only a finite number of nonisomorphic simple $\Lambda$-modules.
\item\label{th:leftnoe} $\Lambda$ is left noetherian.
\end{thmclaim}
\end{theorem}
\begin{proof}
\begin{description}
\item[\ref{th:leftnoe}] Since $\Lambda$ has a finite filtration...\par
\item[\ref{th:radnil}] Let $I$ be an ideal in $\Lambda$...\par
\item[\ref{th:finsim}] If for a $\Lambda$-module $A$ we have...
\end{description}
\end{proof}
\end{document}

And a variation using alphabetical characters:
\documentclass{article}
\usepackage{amsthm}
\usepackage{amsmath}
\usepackage{enumitem}
\newtheorem{theorem}{Theorem}
\newlist{thmclaim}{enumerate}{1}
\setlist[thmclaim,1]{label=\normalfont\textbf{(\alph*)}}
\begin{document}
\begin{theorem}
For a left artin ring $\Lambda$ we have the following.
\begin{thmclaim}
\item\label{th:radnil} The radical of $\Lambda$ is nilpotent.
\item\label{th:finsim} There is only a finite number of nonisomorphic simple $\Lambda$-modules.
\item\label{th:leftnoe} $\Lambda$ is left noetherian.
\end{thmclaim}
\end{theorem}
\begin{proof}
\begin{description}
\item[\ref{th:leftnoe}] Since $\Lambda$ has a finite filtration...\par
\item[\ref{th:radnil}] Let $I$ be an ideal in $\Lambda$...\par
\item[\ref{th:finsim}] If for a $\Lambda$-module $A$ we have...
\end{description}
\end{proof}
\end{document}

And yet another variation changing also the ref key to automatically add "Proof of claim " (this can be redundant):
\documentclass{article}
\usepackage{amsthm}
\usepackage{amsmath}
\usepackage{enumitem}
\newtheorem{theorem}{Theorem}
\newlist{thmclaim}{enumerate}{1}
\setlist[thmclaim,1]{label=\normalfont\textbf{\arabic*.},ref=Proof of claim~\arabic*.}
\begin{document}
\begin{theorem}
For a left artin ring $\Lambda$ we have the following.
\begin{thmclaim}
\item\label{th:radnil} The radical of $\Lambda$ is nilpotent.
\item\label{th:finsim} There is only a finite number of nonisomorphic simple $\Lambda$-modules.
\item\label{th:leftnoe} $\Lambda$ is left noetherian.
\end{thmclaim}
\end{theorem}
\begin{proof}
\begin{description}
\item[\ref{th:leftnoe}] Since $\Lambda$ has a finite filtration...\par
\item[\ref{th:radnil}] Let $I$ be an ideal in $\Lambda$...\par
\item[\ref{th:finsim}] If for a $\Lambda$-module $A$ we have...
\end{description}
\end{proof}
\end{document}

enumerateand then use\labeland\ref– cmhughes Sep 24 '12 at 16:52\begin{description}\item[Proof of claim 3.] blabla \item[Proof of claim 1.] blabla \end{description}– tohecz Sep 24 '12 at 17:20claim1toclaim3because you don't have to bother to think out some names for them. And then if you switch them, you get a complete confusion if you leave the original labels. – tohecz Sep 24 '12 at 18:39