Subjects
Four subjects.
GradeDrive marks Biology, Chemistry, Physics and Maths. Not because that’s all AI can do, but because those are the subjects we’ve tested against real papers and real mark schemes until we trusted the result ourselves. Here’s why those four, and what has to be true of a subject before we’ll add it.
Biology
Structured and extended-writing questions marked against point-based and level-of-response schemes, the same two formats that run through GCSE and A-level Biology papers.
Chemistry
Calculations, equations and defined-answer questions, where a mark scheme states the acceptable working and the acceptable answer, and there's no ambiguity about whether a student met it.
Physics
Working shown step by step, units checked, and level-of-response questions marked against the same descriptors an exam board publishes, not a paraphrase of them.
Maths
Method marks and answer marks, kept separate the way a real mark scheme keeps them separate, so a student gets credit for correct working even when the final answer is wrong.
We only trust a mark if the mark scheme itself can prove it right.
In Biology, Chemistry, Physics and Maths, the mark scheme states its criteria before a script is ever seen: a balanced equation, a method mark, a level-of-response descriptor with its boundaries written out by the exam board. Two different markers checking the same script against the same scheme tend to land on the same mark, not because the question was easy to answer, but because the scheme itself doesn’t leave the mark up to interpretation.
That’s what deterministic means here: the correctness of an answer is fixed by the scheme, not by the marker’s judgement on the day. It’s the property we test for before we’ll add any subject, and it’s why English, Psychology, Geography and the rest stay off the list for now: we haven’t yet proven we can mark them to standards we pride ourselves in.
We only add subjects that we can stand behind.
A subject only goes live once we’re confident we can produce genuinely high-quality results for it: tested against real scripts and real mark schemes, and checked against how a teacher would actually have marked them, not just how plausible the output looks.
That confidence has to come before a subject ships, not after. It’s the same bar every future subject has to clear, not just the four we’ve cleared so far.