Annihilation of Caste, the text that Ambedkar prepared as a presidential address to the Jat-Pat Todak Mandal in 1936 and then published independently when the conference was cancelled, is one of the most uncomfortable documents in Indian political thought. It is uncomfortable not because it says anything that can be easily dismissed, but because it refuses the consolations available to other critiques of caste. The Gandhian position held that caste was a perversion of varna, which had an original form worth recovering. The Marxist position held that caste was a superstructure that would dissolve with the dissolution of economic class. Ambedkar rejected both. He argued that caste was not a derivative of something more fundamental but was itself the foundational structure of Hindu social order, underwritten by religious sanction.
This makes Annihilation of Caste a more radical text than it might initially appear. It is not a reform programme. It is a diagnosis that implies that reform — piecemeal adjustment within the existing structure — is insufficient. If caste is grounded in religious authority, then it can only be annihilated by destroying that authority: by making the shastras answerable to reason rather than placing reason in the service of the shastras. This is why Ambedkar's later conversion to Buddhism was not a departure from his political project but its logical conclusion.
What the text does not offer is a programme for how this annihilation is to be achieved. It is an argument, not a manual. And the distinction matters because it leaves open a question that Ambedkar's critics on the left have pressed: if caste cannot be reformed from within, and if the religious authority underwriting it cannot simply be legislated away, what is the political path?