For A-Levels — extended-response marking
Essay marks that survive a review.
The hardest marking in the system is a borderline 25-marker. Sokros separates what the candidate evidenced from how well they argued it — and writes down the reasoning, so a challenged grade has an answer waiting.
One essay's path to a defensible level.
The same extract → check → judge → record pattern that marks professional qualifications today, tuned for levels of response.
Assessment objectives checked separately
Median-of-runs judging on borderline responses
Calibrated against examiner-marked essays first
01
Read the essay as evidence, not vibes
A 25-mark response is extracted into typed claims — arguments made, evaluation offered, examples deployed — each pinned to the candidate's verbatim words before any judgement happens.
02
Hold the assessment objectives apart
AO1 knowledge credited as AO3 evaluation is how essay marks drift. Each objective is checked and judged separately, so a fluent description can't quietly buy analysis marks.
03
Judge the level, write the reason
The response is placed in a level against the board's descriptors with written reasoning — runnable multiple times with a median taken, so borderline essays get the most scrutiny, not the least.
04
Keep the record for results day
Every level decision and its evidence is stored per candidate. Reviews of marking start from the recorded reasoning instead of a second cold read.
“…however, this depends heavily on the elasticity of demand…” — evaluation present in section 2; counter-argument not developed in conclusion
Around the essay, a working desk.
The same operational layer that runs professional-qualification centres, tuned for extended-response marking.
Borderline essays queue for people
Clear levels resolve on your rules; thin-margin responses route to senior examiners with the recorded reasoning and quotes already attached.
Cohort insights per objective
Where AO3 marks are lost, which questions produce the widest spread, how this year's paper shifted difficulty — read from the record, not sampled.
Integrity for NEA and essays
AI-usage and similarity analysis with flagged passages quoted — including within-cohort comparison — delivered as signals for an examiner's judgement.
Private by architecture
Candidate work stays in your boundary: UK-based, no third-party AI services in the marking path, never used to train models.
Put your hardest essays through it.
Send the borderline scripts your senior examiners debated longest. That's exactly where written reasoning earns its keep.