df = pd.DataFrame({'a': [1, 2]}) df.at[5, 'a'] = 6 print(df) # a # 0 1.0 # 1 2.0 # 5 6.0 

The docstring of DataFrame.at says this should raise a KeyError. This method shouldn't allow the addition of rows because it is ill-performant and makes .at both an indexing and reshaping method.

This may be related to #48224, but seems to me to be a separate issue (at least, from a user perspective).