If the language of thought hypothesis (LOT for short) were right, then the content of a thought would just be the content of the mentalese sentence used by the brain to represent it. I don't see this as a departure from representationalism. It is just one way to cash out the idea that thoughts are the representional contents of representations that are materially implemented in the brain in some way or other. It raises the question who is it who is interpreting the mentalese symbols? If the brain is a syntactic engine, then it does no interpretation. So, the LOT hypothesis threatens to collapse into eliminativism about mental content, it seems to me.

I prefer the antirepresentationalist view that rather equates thoughts (e.g. beliefs, judgments, experiences, etc.) as acts of capacities to believe, judge, perceive, etc. To believe that the cat is on the mat, for instance, is thus an act of representing the cat as being on the mat. It is something that an embodied person does, not her brain. It need not be the representational content of anything. The correct ascription of such mental states to a living individual is grounded on the interpretation of her overall animate behavior (including her linguistic behavior), in such a way that it is disclosed as being rational in light of such a broadly consistent set of mental state attributions. This is roughly Davidson's view of radical interpretation (divorced from his anomalous monism).

This is roughly Davidson's view of radical interpretation (divorced from his anomalous monism). — Pierre-Normand

You remind me of radical interpretation. Great!