Skip to content

Teaching in the portal

Sign in at portal.shuleone.co.ke. Teachers get a tutor view rather than a learner view.

Screen What it is for
Overview Your classes, and what needs attention
Quests The quests your learners are working through
Submissions Work handed in, waiting to be marked
Exams Setting exams, and marking them
Classwork Work set to a class
Marks Marks from portal work
Open learners Learners not attached to a class

CBC Grade 5 science source checks distinguish the reviewed floating/sinking, sound and heat-transfer subsets from practical demonstrations. Written interpretation of supplied records does not certify equipment manufacture, swimming or lifesaving, hearing performance, heat experiments or fire response. The official preview has conflicting summary and detailed organism headings; that mismatch must be resolved before generating a lesson for the affected topic. A source-permission revision needs fresh lesson artifacts; older lesson provenance is retained rather than relabelled.

For England Year 4–6 geography, reviewed lesson subsets distinguish latitude/longitude reasoning, water-cycle explanations and compass/grid-reference methods. A fictional teaching grid practises a method; it does not demonstrate actual Ordnance Survey map reading, real-place knowledge or fieldwork. Four-figure references identify a grid square; six-figure references narrow the area to a 100-metre square when using kilometre grids. They do not establish unlimited point accuracy. Real regional comparisons need verified place evidence before generation. Review each lesson’s stated scope rather than treating one published lesson as complete coverage of a broad geography topic. Selected regional comparisons use verified summaries for the English Lake District, Vesuvius and Yellowstone. Their limited fact bank does not represent a comprehensive study of those regions or authorise real park-management decisions.

Open Content → Create to choose a curriculum level, learning area and mapped sub-strand. Generation checks the exact source, its permitted use and an independently verified set of outcomes for that level and topic. Its quality check must identify where each outcome is taught and practised, and check the tasks and answers. Generated lessons are saved as drafts, including when the automated checks pass. Inspect the finished draft and preview its tasks, answers and media, then use the separate Publish action. Publication runs the required checks again; a passing generation verdict alone does not make the lesson available to learners. An active mapping alone is insufficient. If independent source evidence or permission to generate or deliver content is missing, the studio blocks the operation with a reason. Do not replace the source outcomes with a generic topic to get past that block.

Both the standard quest generator and the newer lesson generator check the exact source version and its generation permission before contacting the model. Missing, disabled or ambiguous permission blocks generation. Permission to generate is separate from permission to deliver a lesson to learners; the normal content and publication checks still apply.

For choice questions, review the question and stored answer key together. An intentionally incorrect learner claim can be a valid unselected distractor when pupils must select correct claims; a question asking for incorrect claims uses the opposite condition. Review support makes stored selections explicit without treating them as proof that a statement is true. Check worked teaching, ambiguous alternatives and answer explanations separately; a passing review still needs inspection of the finished learner experience. Source-review transport guides the reviewer towards actual eligible activity IDs and visible fields. It still permits a failed coverage verdict and missing evidence; exact quotations and publication requirements remain checked by the server. Provider schema guidance keeps nested arrays compact; the server still enforces exact outcome counts, eligible references and quotation limits. This guidance does not guarantee that a reviewer will return valid evidence or identify every teaching defect.

A new review also runs a separate final content check against the complete lesson and its selected outcomes. It checks supplied facts against answer keys and explanations, valid alternative answers, dependencies in ordering tasks, and required applications. A failed, incomplete or unavailable check keeps the lesson as a draft. A draft with an older passing review must be reviewed again before publication; that old verdict cannot skip the new final check. This extra check can still miss errors; it does not certify an entire curriculum or replace checking the finished learner experience.

Questions must supply the visual or explicitly named target they ask learners to use. The diagram-label default displays written clue cards. It needs a configured supported visual extension to show a coordinate grid, circuit or fraction model; a diagram title alone does not draw one. A promised map or compass rose also needs its actual visual. Review the actual preview rather than inferring a figure from the activity name. Literal hemisphere classifications also check that zero latitude is on the Equator, with a suitable boundary category, rather than forcing north or south. An image-description placeholder is not a displayed diagram, and an unmarked number cannot identify an “underlined digit”. For ordering questions, independent steps need an explicit meaningful protocol or a different response format. New checks flag a small set of these defects; teachers should still inspect unfamiliar cases. A released check compares clear, explicit claims about the number of addition regroupings with the supplied arithmetic. Ambiguous cases still need teacher review.

Released on 22 October 2026, both lesson-generation actions create drafts for review and preview. Generation and an AI passing verdict do not publish a lesson; staff must inspect the finished draft and use the separate Publish action after the required checks pass.

For source-backed lessons, a bounded set of well-formed initial review findings can receive one independent recheck against the unchanged complete lesson and exact source. This is not limited to particular contradiction or retraction phrases. Each original finding needs a concrete, quoted resolution: a confirmed defect stays in the review; a finding can be retracted only when the actual question, key or stimulus disproves it. The recheck must also provide a complete replacement review matching the actual source requirements, including the required teaching and practice evidence. Each finding still needs an exact supporting quotation from the questioned activity. A separate visible worked model can supply additional support when the review explicitly identifies that activity and its quotation matches the actual lesson text. Evidence from another activity does not replace the required quotation from the questioned activity. An incomplete or truncated reply, invalid citation, unavailable recheck or unresolved finding keeps the original failed findings held. It does not edit the lesson or bypass any source or publication check.

A separate case allows one recheck when the initial review has no findings and claims complete coverage, but its teaching or practice references are empty. Missing content or invalid evidence still holds publication; a coverage flag alone cannot establish it.

A graded activity can provide both teaching and practice evidence when learners see a complete worked model before answering a distinct fresh task in the same activity. The reviewer must inspect both parts. A brief or case-study is still not practice, and a worked model alone does not prove that learners practised the outcome. These checks can still miss errors, so inspect the finished lesson before publishing.

Drafting instructions may use “present” to retain the initial draft, revised text and concrete changes, but all three outputs must still be requested and assessed.

Source review now keeps a syllabus separate from a curriculum-design framework when their internal numbers match. Each lesson keeps its own source checks. For drafting tasks, equivalent retained outputs such as an original connected passage or sustained comparison and a revised passage or extract are recognised without rewriting a sound task. The response must still retain both versions and explain concrete changes; a final-only response or a promise to revise is insufficient.

Lesson generation now checks English classification instructions against the answer controls: Classify accepts one category for each item. Overlapping memberships need separate multi-select questions or a clearly defined single classification criterion. The checks leave answer keys unchanged. If the same activity repeats this format conflict after one repair attempt, generation keeps an inactive draft for inspection instead of requesting another identical repair. Other content and publication checks still apply. Generation also takes the lesson’s displayed level from its exact curriculum context, so an IGCSE lesson is labelled IGCSE rather than inferred from its numerical demand grade.

Generation guidance separates fresh activities from platform-supplied reviewed models and complete own-response tasks. Where the platform supplies only teaching models, the generator still creates the required fresh own response. Source scope, exact-content checks and publication checks remain in force for the assembled lesson.

Label original fictional records, testimony and interviews clearly as fictional. Attribute wording, data or interviews to real people, institutions or publications only when the supplied source excerpts or pins verify that attribution. Supported real facts and citations remain usable. The generation and review instructions reinforce this rule; teachers must still check attribution in the finished lesson.

In the learner library, a learning-area name can come from the independently reviewed, current source binding. For example, Grade 7 English appears under English even when an older stored subject is named LANGUAGE GRADE7. The original subject and source identity remain intact. An ambiguous Language label alone does not mean English or Indigenous Languages. This naming check identifies the source area; it does not certify that every activity is correct. Matching explicit learning-area names can share one library area. Each lesson keeps its own curriculum and source context: its card shows the actual source credit, or Historical source not identified when no frozen source is known. Sharing an area does not assign a historical lesson to the current reviewed source.

Some lessons address a reviewed subset of a broader official objective. Their checks use the specific lesson targets, while keeping the full source objective and the limits of that subset distinct. For example, a lesson may practise making and interpreting bar charts from supplied tables without teaching dot plots or constructing a table from raw results. A useful published lesson can count toward the starting set; it does not certify that every part of the official objective has been taught. The automatic checks review the stated targets and limits against the exact curriculum source.

England Years 7–9 lessons can use reviewed targets from the shared Key Stage 3 programme. Their year placements are ShuleOne teaching allocations, not a statutory year-by-year sequence. Check the selected year and topic before generating.

England Years 10–11 source checks require the exact reviewed Key Stage 4 edition, current topic ownership and complete source actions, alongside separate source permissions. These are selected targets from the shared KS4 programme, not a statutory annual allocation or exam-board qualification. A target needing an unapproved original diagram or graph remains blocked before generation and publication; a source mapping alone does not make it available. Interpreting a supplied graph does not certify that a learner constructed one.

Twenty selected topic placements across England Years 10–11 now have reviewed source permissions for Mathematics, English and Science from the Department for Education’s December 2014 programme. Only the exact reviewed targets are eligible; required model or asset approvals and each lesson’s own checks still apply. Source admission does not publish lessons or certify learner completion. For electricity, the reviewed starting subset uses supplied current, resistance, charge and power records and series-resistance reasoning. It does not certify domestic a.c. supply, mains-wire or full electrical-safety coverage, or actual circuit construction.

For the reviewed right-triangle, time-series and atomic-structure models, keep the required figure, equivalent table and complete worked explanation together, before the fresh questions. Studio reports missing or changed model content before publication. The illustration control cannot redraw these required models. Restore the reviewed model when a diagnostic identifies a change; an approval label alone cannot replace its visible teaching. Learners must still answer different questions in their own words.

The selected Cambridge Lower Secondary Stage 7 cell and particle-model topics have separate exact-edition, model and source-permission checks. Keep their reviewed original figures, equivalent text tables and completed examples together before fresh practice; these protected models cannot be redrawn. The combined nonfinal written task must perform every selected target, including constructing the requested fresh particle representations with a usable legend. Model presence alone does not prove learner performance. These checks do not activate source rights or make the whole Stage 7 programme available; normal source permissions and lesson review must still pass.

The CBC Grade 5 preview pilot is limited to ten reviewed topics across Mathematics, English, Science and Technology, and Social Studies. Only the admitted targets are eligible after their source permissions are enabled. A public preview is evidence for original lesson authoring; it does not permit copying the source PDF. Other topics remain blocked until their own source checks are complete.

The separate CBC Grade 6 pilot also covers ten reviewed topics in those four learning areas, after its own source permissions are enabled. Check that learners actually create their multiplication patterns, reading-advocacy messages, school–community strategies and peace-promoting responses where required. Recognising a supplied example does not replace creating one. Supplied observations and fictional peer viewpoints assess a bounded lesson task; they do not certify physical practical work, real collaboration or enduring attitudes. An earlier lesson’s published status alone does not certify these reviewed targets.

Grade 6 English Guided Writing: Filling and Designing Forms uses complete fictional records. Learners fill and design forms without supplying their real contact or family details. The previous 1.4.1 Mechanics of Writing path is retained with its historical results and excluded from current source admission; those results are not evidence of form-design competence. Kiswahili Vinyume vya Vivumishi requires learners to create sentences with correct agreement, as well as recognise adjective opposites in context. The stable source code KISWAHILI_LANGUAGE resolves to Kiswahili when its subject name corroborates it; conflicting languages still fail. Source permissions and exact reviewed-topic evidence remain separate requirements.

Reviewed England Year 4 subsets distinguish written word-meaning and spelling work from oral reading and handwriting performance. Science subsets assess digestion, teeth, food chains and written circuit reasoning; they do not certify a physical circuit build. Availability still depends on a lesson passing publication review. Additional Year 4 English practice can rehearse follow-up questions, justified replies and responses to different viewpoints in writing. Typed answers do not assess actual listening or spoken fluency. Scientific-enquiry practice uses complete supplied melting records within states of matter; it does not certify that the learner performed an experiment or took measurements. These remain selected source subsets, and lessons still require publication review.

For story predictions, judge the evidence available before the ending. Several future actions may be plausible: a character trait alone does not establish one uniquely correct prediction, and the later ending cannot make an earlier reasonable prediction wrong. A narrative sentence mentioning something a character watches does not itself instruct the learner to view a video. Genuine unsupported viewing instructions are still held. When reviewed source evidence requires a combined writing task, it comes before the final consolidation and must collect every selected action, including original production where required. If an otherwise grounded passing review omits or duplicates activity checks, the engine may request one new review of the complete unchanged lesson. It never fills missing checks itself. Substantive failures, unresolved checks and unsupported quotations do not qualify for this recovery; an incomplete second review stays held. Case-study cards are read-only worked records. Authoring now holds a case-study with hidden choice keys or writing rubrics: use a supported choice, data-reasoning or written-response activity to collect an answer. A rubric stored behind a reading card does not establish that the learner produced a response.

New final-review requests give every activity its own required named slot. This avoids counting or repeating array entries. The server preserves the original response and checks every slot, verdict and citation before converting it to the stored activity list. Missing or extra activities, duplicate JSON names, invented citations and unresolved judgements remain held. Existing complete review evidence stays supported. When an initial review raises a choice-related finding, its independent adjudication now states which actual option is involved and whether it is accepted by the key. An intentionally incorrect unkeyed option alone cannot confirm a teaching defect. An alleged missing key must name an option actually offered; a select-from-list question need not list every member of a wider curriculum category. Invalid or unresolved evidence stays held. Retraction still requires complete source review and the final independent content check.

Reviewers must distinguish intentionally incorrect or nonmatching options from endorsed teaching. They check the displayed wording, actual task, stored answer key and feedback; internal activity and option IDs do not add learner-facing meaning. Any valid answer rejected by the key remains a publication defect.

Additional Cambridge Stage 4 Mathematics subsets use the exact registered 0096 framework for fractions/percentages, time, rectangle/square formulas, compass descriptions, statistical comparisons and chance vocabulary. Their full source context and unassessed practical components remain visible in source scope. Admission does not itself publish a lesson or establish complete mastery of the wider framework topic.

Additional Stage 4 English subsets use the exact registered 0058 framework for selected punctuation interpretation, paragraph structure, evidence-based reading, personal written responses, possession/agreement, original short texts, alternative story openings/endings and proofreading. Their source scope states the selected actions and the wider objectives that remain unassessed. Typed work does not certify spoken fluency, handwriting speed, sustained reading habits or the truth of a self-reported personal experience. Publication still requires the normal reviews. Reviewed class and subject bindings must also match the requested topic; a mismatch stays held.

The additional Stage 4 Science subsets use the exact registered 0097 framework. They cover selected written reasoning about models, skeletons, energy for life, habitats and food chains, particles and materials, energy transfers, light, circuits, Earth and space, and scientific evidence. Each topic retains its full source objectives beside the selected targets. These digital lessons do not certify practical construction, measurements, experiments, medical knowledge or other actions outside their stated scope. Original teaching and questions remain subject to source permission and normal publication review.

Grade 7 English Polite Language and Handwriting: Legibility and Neatness include bounded digital lessons. Polite Language assesses recognition and written rehearsal of courteous introductions; it does not certify speaking or respectful behaviour in a real exchange. Handwriting assesses knowledge of readable writing and its communication benefits. Actual handwriting production needs paper or authentic ink/image assessment; typed answers do not establish neat handwriting. Optional paper practice is not automatically marked as completed. The current Social Studies lesson is Peaceful Coexistence, under People and Relationships; the obsolete Peace and Conflict Resolution catalogue path remains excluded.

For the prepared CBC Grade 7 topics, Studio distinguishes the Reviewed source heading from the retained Catalogue label used for grouping. The editor shows the same distinction. If the exact source heading cannot be verified, Studio displays Reviewed source heading unavailable; the catalogue label is not a substitute. These display checks do not enable Grade 7 authoring or learner delivery by themselves: the exact source bindings, permissions and current lesson checks must also qualify.

The prepared CBC Grade 8 pilot is limited to ten exact topics across Mathematics, English, Integrated Science and Social Studies. It keeps 43 full source actions as context and assesses only 39 bounded lesson targets; four practical or relationship outcomes remain held. Source permissions and the current source mapping must qualify separately. Existing lessons and catalogue names are not proof of this new source scope.

The cell and map topics require their reviewed original images and complete supplied fact records. Generation stops before an AI call if the required image cannot be verified. Later edits must retain the full visible model. The supplied lens totals and Plot P area are worked teaching: learners need a fresh calculation with different settings or dimensions, alongside their own written reasoning. These tasks do not certify microscope operation, actual cell observations, fieldwork or lasting attitudes.

For signed integer number-line tasks, use a supplied approved fixed-scale image with all coordinates and equal unit spacing explained in visible text. Learners can type their start coordinate, direction, number of unit steps and endpoint using the existing working controls. The chart tool uses a zero baseline and nonnegative values; it is not an interactive signed number line. Give a worked model before a fresh task with different values. This guidance does not add an asset to a source’s required image list or certify free plotting.

The reviewed CBC Grade 9 source subset covers ten exact topics in Mathematics, English, Integrated Science and Social Studies. Its 46 full source actions remain context for 40 bounded targets; six whole outcomes and the oral component remain held. The installed engine checks the exact reviewed source edition, topic, current mapping and independent-authoring permissions before generation or delivery. Lessons still require content review and student-side audits. Use complete supplied records and fresh learner production; this subset does not certify calculator operation, electron drawing, water softening, oral performance, community mediation or lasting attitudes.

The Cambridge Stage 6 pilot supports original lesson authoring for ten reviewed topics: three Mathematics, three English, three Science and one Computing. Its lesson targets are drawn from 26 selected official objectives. Check the selected topic and its stated limits; this pilot does not certify complete Stage 6 coverage, anatomy, flowchart construction or physical practical work.

Where required, learners must derive a sequence rule, write sentences using parenthetical commas, dashes and brackets, produce an argument with a plan, draft and meaningful revision, or create and test a two-variable algorithm. Selecting supplied examples does not replace that work. Complete local passages, models and instructions must support the task. Source admission alone does not mean that ten lessons have been generated or that their learner audits are complete; each lesson still follows the automatic source, task, answer and narration checks described above.

To build a starting set across learning areas, select the class and choose Fill class to 10. The preview shows how many uncovered topics it will generate to reach ten published V2 lessons. Existing published lessons count toward ten; drafts do not. Topics with an existing V2 lesson are skipped, and the new lessons are spread across learning areas. Keep the page open while the batch runs. If a lesson fails a check, repair its draft and run Fill class to 10 again for the remaining places.

Open Content → Library to see the published lesson. If a quality check or narration fails, the lesson remains a draft with its issues shown in the studio. Use Edit to inspect the learner activities and answers, then regenerate or retry publication after the issue is resolved. A draft does not appear to learners. The source version and curriculum identity remain attached to every lesson.

Tables need clear column labels and a matching cell for each column in every row. A lesson with an invalid or unlabelled table stays a draft. Use Edit to repair the labels or cells, then review the corrected table before publishing. Intentional blank form cells can remain blank.

For single-choice table questions, ask about one clearly identified item, or offer choices that each give the complete requested mapping. Do not ask learners to classify every row when each choice answers only one row. Review checks flag this mismatch; inspect the task and every option together.

For the reviewed Grade 4 subtraction and multiplication lessons, and Grade 5 addition, subtraction and multiplication lessons, learners must create their own four-term pattern, show all three transition equations, and explain a linked practical use in one written response. Grade 5 addition and subtraction terms stay within one million. The Grade 5 growing multiplication task specifies a multiplier from 2 to 9 and terms no greater than 1,000; follow the bounds shown in the task. Choosing or ordering a supplied pattern is practice; it does not replace this creation task. A missing or changed required task holds publication, and an outdated task review makes that lesson unavailable until repaired and reviewed.

CBC Grade 10 has separate 2025 reviewed pilot source identities for Core Mathematics, English and Chemistry. Match the school subject to the exact source and version when configuring its curriculum. The earlier source identities remain separate. A source selection does not grant lesson-generation or delivery permissions; those checks still apply before a lesson becomes available.

For the prepared Core Mathematics update, Real Numbers, Indices and Logarithms, and Quadratic Expressions and Equations each require 12 activities. Set Activities per lesson to 12 activities. Leave Required specialist activities empty, or select Handwritten working; the other specialist choices do not fit these reviewed lesson sets. An incompatible count or format stops the request before paid generation starts. Keep the required worked models and learners’ own production tasks. The quadratic lesson must retain its reviewed area image, dimensions and supporting text; asking learners to imagine or draw an unseen model does not replace it. These prepared checks do not announce availability: source permissions and current automatic lesson checks must still qualify. Learners’ written responses continue through the usual marking process.

The prepared England Year 6 Animals including humans lesson has a fixed set of 10 activities. Set Activities per lesson to 10 activities and leave Required specialist activities empty. An incompatible count or format stops before paid generation. Keep both reviewed circulation and absorption models, their complete local explanations, and the one task asking for the learner’s own linked explanation. The lesson uses a simplified human functional model, not an anatomical drawing or a practical or medical assessment. It distinguishes lifestyle support and risk from a guarantee of health, and does not treat all medicines as harmful. Written responses still use the usual marking process; the fixed task checks do not certify a learner’s understanding. Changing or omitting a required model, task or answer binding produces a specific issue. Restore the reviewed task before re-review; do not substitute an unseen diagram or a copied completed answer. This prepared pilot does not announce learner availability.

Ordinary tables supplied with an activity show their column headings, all rows and caption before the task, including multiple-answer questions. On a narrow screen, swipe the table sideways to read the remaining columns. Each row must match the column count; cells use plain text, numbers or an explicit blank, so hidden objects cannot replace evidence.

Evidence-matching records show a source label and its detail. Put the full evidence body in detail so learners can read it. Source labels must be nonblank text; optional detail and text bodies must also be text when supplied. Leave detail out or use an empty string when the label contains all the evidence. A hidden or conflicting text body produces a validation issue; a source label alone remains valid when it contains the evidence.

Case-study observations must contain a text label and visible text in detail or value. A sentence saved without those fields cannot form a readable observation row. Correct the reported row before reviewing the lesson again. Classification activities must give each item and each category its own distinct ID, with readable text labels. Each item’s answer must refer to one of that activity’s category IDs. These checks apply across curriculums and prevent blank records or questions with conflicting answers.

Scored ordering activities need one defensible order. Do not split simultaneous actions into separately ordered cards: for inhalation, group diaphragm and intercostal muscle contraction into one action before chest volume increases, pressure decreases and air enters. English lesson checks now block the known separate-muscle ordering pattern; a grouped process or a non-ordering question can pass. This bounded check does not prove that every other sequence is correct.

For handwritten working, each required step represents one required result or calculation. Every accepted response for that step must be a complete alternative, not another required part. For example, put moles of K and moles of O in separate required steps, each with its own acceptable working. The recognition context describes the writing format; it does not set the learner’s curriculum or grade.

For source-backed written responses in the selected canonical lesson sets, marking can use the exact preceding reviewed teaching models paired with that response task. The fresh question and its supplied records still determine the current case: a worked example’s names or numbers cannot replace them. Unrelated teaching, later activities, hidden answers and post-answer feedback are not added as evidence.

If the checker cannot verify every part of a response, it remains unmarked. The message now distinguishes this uncertainty from a service failure. Your answer is kept, and this does not count as a wrong answer; uncertainty does not award a pass either. There is no automatic marking retry.

For pilot lessons, availability requires the exact reviewed topic and source, current automatic lesson checks, and permission to use that source. If the source or required evidence changes, an earlier lesson may become unavailable. Valid regenerated or re-reviewed content must pass the current checks before learners can open it again.

The reviewed Stage 4 Science object-key lesson uses a fixed set of four described shape-and-colour objects. Its eight activities show the complete key locally and check both decisions on a route. Changes that contradict the supplied observations, branches or answer bindings keep the lesson as a draft. This lesson practises applying that closed key; it does not certify identifying materials or living things.

In Content Studio, single generation and batches save drafts for review. The automatic Publish after checks option has been removed. Use the Drafts filter to find the results, inspect the complete lesson and preview its audio and visual material, then use the separate Publish action when it is ready. Missing, pending or unavailable quality results still need attention and do not make a draft ready to publish. Generation, source, model, narration and illustration checks still apply; generating a draft can still incur media costs. Each illustration run makes at most three picture attempts, including failures. Existing pictures and empty drawing briefs do not use those attempts. A failed generation request is reported instead of switching to automatic publication. Draft status and passing checks do not certify that every answer is correct.

Released on 21 October 2026, generation assembles the exact reviewed teaching models automatically for the selected Stage 7 Science, Mathematics, English and Computing topics and the three figure-protected KS4 topics. Where a complete reviewed written task exists, it includes that task too. The requested activity count includes these supplied blocks; the generator creates the remaining fresh practice around them. Models stay before practice, and a supplied written response stays before the final consolidation. Each supplied block is checked against the full activity format before generation starts. Missing presentation labels receive fixed teaching or response labels; the reviewed content remains protected. A block that still fails those checks stops generation.

This applies to new generation. Editing a lesson still requires preserving its protected models and full task parts. Extra or altered protected blocks returned during generation remain validation failures and can be inspected in the saved draft. Source permissions, full lesson review and publication checks still apply; publication requires the separate Publish action after staff review. This support does not activate an entire curriculum or certify that a lesson is correct.

Chart checks now reject an explicit power-of-ten axis scale that disagrees with the declared unit. Chart values and answer options use that declared unit; the system does not silently rescale them. These checks catch supported notation mismatches, but do not replace checking the chart’s meaning and answers.

Some numerical single-choice checks now catch equivalent keyed answers and ambiguous positive standard-form choices within supported notation. Explicit answer-format constraints still matter. For the twenty reviewed KS4 subsets, review evidence must link the same combined written response to every selected target; separate calculation practice remains allowed. These checks do not prove that all teaching, answers or source coverage are correct.

New reviews of the twenty selected KS4 subsets also require a short exact citation from visible teaching for every referenced teaching step. Completed models must be visible before learners respond; hidden accepted answers and feedback shown only afterwards cannot supply that evidence. Supported text includes passages, observations, worked step labels and string cells in ordinary tables. Numeric-only table cells and figures need supporting visible text in a supported field. Repair the actual teaching before reviewing again; do not invent a quotation or just change its reference. These checks verify the recorded text links, not whether an explanation is complete or correct. Reviews match quotations to the teaching text as displayed. Formatting differences are handled only when the full quotation still matches that visible text; changed or unbacked quotations block publication. These checks do not certify teaching quality.

Existing reviewed revisions keep their current gates; a new substantive revision receives the new evidence checks. New generation saves drafts; publishing is a separate action.

If repeated generation attempts fail validation, the studio keeps the closest editable candidate as a separate draft and lists its specific errors. Open its draft ID or Edit to inspect and repair it. It counts as a failed generation, not a published lesson, and does not replace an existing published lesson. If no editable candidate can be saved, the studio reports that explicitly.

A missing teaching or practice reference now identifies the affected outcome index and activity IDs. Each selected outcome needs both teaching and a learner-response task; a worked case study alone supplies teaching. Repair the actual missing task and its references, then review the draft again. Changing a coverage flag or relabelling a case study does not supply learner practice. These messages explain the recorded evidence gap; they do not certify that every other activity or factual claim is correct. A missing reviewer coverage record now names its exact outcome index, including when other records are duplicated or out of range. Inspect the existing activities and re-review their actual references before adding or changing teaching: an omitted review record alone does not prove that the lesson content is missing.

For one-category-per-item questions, use one stated classification dimension and make sure each supplied item has only one valid category. A valid alternative answer must not be rejected merely because a different valid answer was saved. Required teaching must include actual explanations and worked examples, rather than only goals or answer criteria. Automated review is instructed to flag these defects; inspect the full draft and learner experience because a passing review is not a guarantee of correctness.

For multiple-answer questions, instructions must tell learners to select every applicable answer. Some explicit English “choose one” instructions are blocked when more than one answer is keyed; correct the task wording without dropping valid answers. This bounded check does not interpret every instruction or language. For grammar revisions, supply the intended actors and timeline, accept clear meaning-preserving alternatives (including valid singular they), and retain justified changes of time rather than forcing every verb into one tense. Review still requires checking the actual task and answers.

Choice feedback must identify the answer by distinctive wording or content, then explain why it fits. Displayed choices can change order and letters, so avoid feedback such as “the first option” or “the first opening” when choosing between supplied openings. Some explicit English position references now block publication; correct the explanation without changing valid answer keys. Genuine source labels, passage order, times and procedural sequences keep their meaning. This bounded check does not interpret every reference or language; review the complete task and feedback together.

For chemical-equation drafts, check each element against the actual formula in the worked model, answer guidance and feedback. A coefficient multiplies the whole formula; changing a subscript changes the supplied substance. Use formula-unit or ion-ratio language for bulk ionic solids, while retaining correct molecule language for molecular substances such as water. A balanced equation checks atom bookkeeping; it does not prove that a reaction was performed or establish its conditions, speed, yield or safety. Generation and automatic review include this guidance; a full factual check is still required.

Supply the actual complete worked text when asking learners to read a model. Generic statements such as “the text begins with an introduction” describe a text but do not provide it. A bounded check holds this placeholder pattern, including “read the following model text” followed only by descriptions of its opening, middle and conclusion; reviewers must still check all teaching records and fresh tasks in full.

An explanation may define the keyed word and separately contrast each distractor. This explicitly scoped feedback is allowed; simply naming or endorsing every option still requires correction, and semantic review must check the actual reasoning.

For supported literal “select all correct conversions” tasks, an exact arithmetic check now compares every offered standard metric/time relation with its key. Correct alternatives must all be selected; incorrect relations must be excluded. This bounded English check does not interpret approximate, compound, squared, offset or author-defined units,or tasks with extra selection constraints. Those still need full semantic review.

A reviewer’s retraction must still include a complete replacement findings list and source coverage. Supported review providers are now asked for that structure directly. A missing field, incomplete response or refusal keeps the lesson in draft; it cannot become approval by removing the field check. A valid output format also does not prove factual correctness.

Science checks distinguish safety risks from controlled variables, solubility from dissolving speed, and simulation outputs from physical observations. Repeated measurements can expose variation without proving instrument accuracy. Compare complete supplied conditions and contextual features; broad claims such as every bird flying or a seed always germinating need correction. Teaching and assessment remain limited to the selected source objectives.

A written calculation may mix symbols and words, such as “6/10 = 6 divided by 10 = 0.6 = 60%”. The bounded literal checker now leaves that unsupported word-operator chain to semantic marking instead of cutting it into false partial equations. This does not approve the whole explanation or waive its required evidence.

Supported neutral English decimal/percentage recognition now checks every offered value exactly: 0.3 = 30%, while bare 30 has a different value. Percentage-only instructions retain their representation requirement. Bounded duration checks also hold competing single-choice answers such as 95 seconds and 1 minute 35 seconds, and incorrect or artificially forced shortest-to-longest duration orders. Unknown constraints, approximations and unsupported wording still need semantic review. Promised triangle-part and shaded-fraction figures must actually be rendered; generic clue icons do not supply that geometry.

A worked arithmetic example must show its complete teaching before asking for fresh working. A response control whose accepted solutions are hidden until checking cannot serve as the only model. A bounded check now holds that captured pattern. Reflection, translation and plotting tasks also need an actual rendered figure; generic clue icons and a diagram title do not supply a coordinate grid.

Automated lesson-audit workflows can request a text-first draft before preparing preview audio and optional pictures. The normal Content Studio Generate action keeps its preview media behaviour. A text-first draft still undergoes source and content checks; explicit publication reruns the current content review and prepares required narration. Optional pictures can be added through Illustrate. This avoids preparing media for text that is about to be corrected.

Coordinate construction now supports all four quadrants when a lesson supplies a negative grid minimum. Equal-unit axes cross at the true origin; reflections and translations can plot negative coordinates using the existing keyboard-accessible choices. Lessons without a negative minimum keep their first-quadrant grid. Point labels below zero are placed clear of the horizontal-axis numbers. Digital plotting does not certify physical drawing precision.

England Year 6 maths has six further source-reviewed subsets for formal methods, ratio and sharing, conversions, angle relationships, four-quadrant transformations, and supported line graphs with the mean. Each states its limits; the statistics subset does not claim pie-chart construction. Normal content and publication checks remain required.

England Year 5 maths has additional source-reviewed subsets for formal addition and subtraction, formal multiplication and division, metric/time conversion, angles, and reflection/translation. Lessons state their selected limits. Geometry tasks must use actual on-screen construction controls; typed answers do not establish physical drawing precision.

Stage 6 English also has selected source-reviewed reading, grammar, text organisation, response, spelling, vocabulary, extended writing and revision subsets. These are matched to their exact Stage 6 class, subject and framework context. Wider objectives and oral or handwriting performance remain outside each lesson’s stated scope.

Stage 6 Cambridge maths also has selected source-reviewed right-triangle area, integer four-quadrant transformations, and probability/event-overlap subsets. Fractional coordinates, diagonal reflection, rotations, physical construction and conducted chance experiments remain outside those selected lessons.

Stage 5 Cambridge maths has additional source-reviewed subsets for signed sequences, decimal place value, fraction/decimal/percentage connections, time intervals, supported coordinate transformations, and likelihood language. These retain the complete source-group context and explicit selected limits; a supplied probability model does not certify that a learner conducted a chance experiment.

Stage 5 Cambridge science has fourteen further source-reviewed subsets spanning enquiry, plant reproduction and interactions, material properties and changes, selected forces and sound concepts, magnetism, Earth/space and evidence-based science discussions. Every subset preserves its complete framework group and states what it assesses. Supplied records and simplified simulations do not establish that a learner handled equipment, measured accurately, conducted a real experiment or researched their own local area. The water-loss simulation uses a compressed virtual timer and chosen model rates; its results are not physical measurements.

Stage 5 English has additional source-reviewed subsets for speech interpretation, text organisation, personal response, spelling, vocabulary, composition and proofreading. Each lesson states its selected scope; typed responses do not establish oral performance, handwriting, sustained reading habits or every objective in a broad source group.

For error-identification questions, an erroneous sentence can be a correct selection: review the learner’s response against the actual instruction. A review hold can be retracted with quoted task evidence and the actual keyed option; the independent source review and final blind content check must still pass. This does not automatically approve keyed answers.

Check every answer option under the question’s stated conditions. Equivalent expressions such as 2(L + W) and (L + W)2 are both valid factorised forms; one cannot be a wrong option solely because its factors are reordered. A stated requirement such as expanded form can distinguish an equivalent expression that does not meet that form. State restrictions such as a nonzero denominator when needed, and preserve them when simplifying. Replace an equivalent distractor or choose a format that accepts every valid answer.

For unqualified fraction-value questions, 4/12 and 1/3 are the same answer. The shared draft and publication validation blocks this ambiguity in supported plain fraction questions. A request for a specific representation, such as lowest terms, needs separate review; this bounded check does not decide whether a question’s key is mathematically correct or cover every word-problem format.

When a lesson classifies sources, state the enquiry and enough origin/context to explain each role. A later first-hand recollection can still be primary evidence for that event; age or source type alone does not settle the role. Sources and their agreement do not guarantee truth: learners should consider origin, purpose, perspective, limits and corroboration. In evidence matching, each whole claim needs one defensible source under its exact wording. Narrow an ambiguous claim or use an appropriate multi-answer or reasoned response; one source may support several different claims.

Provide visible worked teaching before a fresh response needs the skill. Goals, key terms, answer criteria and feedback shown after checking do not replace that teaching. Put the complete relevant passage or data in the activity that needs it, so learners can answer without reconstructing missing evidence from memory or navigating back. A worked model includes the completed operations, reasons and final result. An unanswered question or an expression left for learners to solve is a task, not a worked solution. Use different given values for the later fresh task. Case-study givens saved in the former hidden scenario field now move into the visible passage when no passage exists; conflicting passages need correction rather than replacement.

The total lesson estimate must equal the sum of all activity estimates, in positive whole minutes. Include reading and reflection in the relevant activity estimates; this format has no separate unallocated-time field. The editor reports a mismatch for repair, and review or publication waits for consistent estimates. Completion headings describe practice completed, rather than claiming mastery, expertise or certification from one short lesson.

For newly generated lessons, ShuleOne calculates the total from valid activity estimates. Check that each activity estimate is realistic; generation does not change those estimates. Invalid estimates still need repair. When editing a lesson manually, keep its total consistent with the activity estimates before review or publication.

Sequence and ranking activities order every supplied item. Their answer must list every item ID exactly once, with no unused distractors. For missing-term questions that offer extra options, use a choice format instead. A reported missing, duplicate or foreign ID needs an authored correction; do not invent a full order or silently drop an option. A scored order must follow a stated objective basis, such as supplied measurements, times, procedural dependencies or language composition. Values, attitudes and subjective importance do not have a universal correct rank; ask learners to explain their own reasoning instead. A valid permutation alone does not establish a defensible order.

For a rejected open response outside mathematics, the feedback now points back to one visible writing-guide entry when an entry can be selected safely, or asks the learner to check every part of the question. It preserves the marking verdict and avoids adding unverified praise or corrections. This checking cue does not prove which criterion failed; learners can still use Ask tutor for further help. Accepted answers retain their usual feedback, and mathematics retains its calculation-and-units checking cue.

The Grade 4+ lesson tutor checks supported explicit numerical equalities in fresh and saved replies across learning areas before displaying or narrating them. A false or undefined equality, or a reply beyond the check’s bounds, receives a general checking hint without caching the unsafe response. The check is conservative: quoted false examples can also trigger the hint. Unsupported notation and claims without an explicit equality still require judgement; this does not certify the whole reply.

Tutor requests include the learner’s complete bounded answer and supported false or undefined literal calculations as prompts for contextual inspection. The tutor is asked to address a current calculation error before requesting reasoning already present, while respecting quoted mistakes, retractions and approximations. The first hint protects the answer; this adds no marking verdict or extra AI request and does not guarantee a relevant hint every time.

Written-answer checks also recognise a bounded set of explicit English statements pairing a nonnegative digit value with a named decimal column. These use the exact column value while retaining contextual checks for quoted mistakes and corrections. An approximate label on a false exact column value stays unmarked. Signed-value claims and other prose remain outside this mechanical check; learners do not need a new answer format. This does not establish correct digit positions, scaling reasoning or mastery. The lesson tutor also withholds recognised false column-value statements using its existing safe fallback, including some quoted false examples. Immediate numeric bracket products are checked with the same exact arithmetic and context rules. Ambiguous division mixed with an implicit product, spaced or symbolic products, and ambiguous grouped-number prefixes remain outside this bounded check. It does not validate a sequence rule, algebraic model or every prose assertion.

For an authored place-value response that requests labelled original numbers, keep each number declaration and its digit-value calculation together. Use the displayed labels, digits 0–9 and calculation format; identify the place explicitly when the chosen digit repeats. These tasks check supported whole-number digit values against the declarations alongside the response criteria. If the required evidence cannot be verified, the answer stays unmarked for clarification and another attempt. Keep the complete response within 2,000 characters. This additional check applies only to tasks that explicitly request it; it does not independently verify every mathematical assertion or establish mastery.

Non-mathematical written explanations use the rendered question as well as displayed passages and tables for structured criteria and material-claim checks. A proposed acceptance receives one separate same-client assertion check, including recalculation of material quantities. It receives the original task and answer, rather than the first marker’s verdict or selected claims. A correct point does not cancel an unsupported or contradicted claim elsewhere; an example explicitly corrected by the learner differs from an endorsed claim. Missing or ambiguous checks leave the answer unmarked for retry without a wrong-attempt score. Personal views and creative choices do not require one fixed interpretation.

The saved numerical lesson score includes only activities with deterministic answer keys. Open written responses and confirmed working receive formative checks separately; an accepted response does not add points to that score. A full numerical score therefore does not certify the learner’s written reasoning or independent mastery.

These model judgements can share errors. Evidence coverage, spans and structured consistency do not prove semantic completeness or accuracy or replace teacher review. Hidden keys and source or observation arrays that are not displayed are not learner source evidence. Keep complete responses within 2,000 characters.

If the source release has changed, generate a new lesson for that topic. The studio checks the current release before generating and keeps the earlier draft and its source record for inspection. Retrying a review on an outdated release stops before an AI review is charged. A valid new lesson still follows automatic publication checks.

Saved lesson counts can remain visible for resolved subjects while another subject’s source is unavailable. A partial-update notice does not mean those saved completions were lost. Check the affected subject’s curriculum source in the school settings; retrying cannot repair a missing or conflicting source identity. Exact Kiswahili sources resolve automatically when their subject and language identities agree. This does not approve their lesson content or source permissions.

Selected Cambridge Stage 7 lesson preparation

Section titled “Selected Cambridge Stage 7 lesson preparation”

The prepared Mathematics, English and Computing support covers selected reviewed objectives only. It does not make an entire subject or framework available. Source permission, activation and each lesson’s publication checks remain separate requirements; a prepared topic can stay unavailable until those requirements pass.

For these selected topics, keep the supplied original completed models before fresh practice and preserve their full passages, tables and explanations. The combined written response must require every selected objective and retain all supplied records and task parts. A completed teaching example is not evidence that the learner has practised it. Review references must point to the visible teaching and the learner’s actual task.

The prepared Computing tasks use manual program traces and supplied data tables. Learners explain the program’s inputs, variables, operations and outputs, or reason about identifiers and queries from the supplied records. A typed trace does not show that software ran or that a live database query was executed. These bounded tasks do not certify practical software skills beyond the reviewed objectives.

Selected Cambridge upper-secondary lesson preparation

Section titled “Selected Cambridge upper-secondary lesson preparation”

Prepared support covers 30 selected topics across Mathematics, English, Biology and Computing at IGCSE, AS Level and A Level. Availability still depends on separate source permission, activation and publication checks. Other topics and complete syllabuses are not included by this preparation.

For each selected topic, generation keeps every supplied completed teaching model and one complete combined written task. These blocks count towards the normal eight-activity lesson; separate fresh final practice remains required. Keep every requested part in the written response, including full initial and revised texts with comparisons where asked. The response box supports up to 6,000 characters.

Computing examples use complete authored programs, manual traces and supplied tables; they do not record software execution or live database results. Biology examples use supplied teaching models rather than observed experiments or clinical guidance. Numerical iteration tasks state a finite stopping row and a separate accuracy check; they do not require 80-digit arithmetic or software execution.

Selected CBC Grade 11 and 12 lesson preparation

Section titled “Selected CBC Grade 11 and 12 lesson preparation”

Reviewed support covers ten selected topics at each grade across Core Mathematics, English and General Science. Generation keeps complete original teaching examples and one fresh written task covering every selected target. These blocks count towards the normal eight-activity lesson, with separate fresh final practice. Source permissions are enabled for these reviewed subsets; each lesson still needs its normal content and publication checks. The whole courses are not included.

The examples use written explanations and supplied records, calculations and intended meanings. They do not certify practical experiments, electrical work, clinical decisions, software execution or observations the learner has not made. Complete every requested part of the written task. Public curriculum previews establish the reviewed edition and selected targets; they do not grant permission to reproduce the publisher’s materials.

The prepared review update checks actual writing-process evidence when the selected objectives require drafting and revision. Before fresh practice, supply a complete original source pack, a completed plan for the audience and purpose, the initial draft, the revised text, and specific changes with reasons. A title, generic feedback label or instruction to revise cannot replace those examples.

The learner’s response must retain the plan, initial draft, revised text and a comparison of concrete changes. One answer box can hold labelled parts. Include these parts in the task instructions and marking requirements; a polished final text alone does not show revision. The retained-output check is limited to explicit supported target wording; the full lesson still needs review for teaching quality and source coverage.

Source resolution recognises the exact Cambridge First Language English 0500 and English Language 9093 identities. Conflicting or unspecified language identities still block generation, and source permissions remain separate.

Choice-review guidance distinguishes deliberately incorrect, unselected learner claims from errors in teaching or a stored key. Authors should still inspect the actual question, alternatives and feedback before accepting the finished lesson.

  • Prepared writing-process checks require retained drafts and a meaningful revision comparison where selected objectives call for them.

  • Selected Stage 7 science tasks now protect their exact reviewed teaching models and require an original combined response for every selected target; source permissions remain a separate gate.

New reviews of the twenty reviewed KS4 subsets check short quotations linking each teaching reference to supported visible text. Hidden answers and later feedback cannot serve as worked teaching. Existing reviewed revisions retain their current checks; new substantive revisions use the new evidence format.

The reviewed England Years 10–11 starting scope now has source permissions for twenty selected topic placements. Check the exact target and any required model approval before generating.

KS4 admission checks now keep source identity, current topic ownership, permissions and required asset approval separate. Biological transport generation and review guidance also distinguishes blood routes, airflow and separately carried substances; it does not turn one oxygen molecule into a carbon-dioxide molecule. These checks and model reviews do not guarantee that every scientific error is detected.

Question-only source records now join displayed passages and tables in source-aware written checks. Proposed non-mathematical acceptances receive a separate assertion check; unavailable checks remain unmarked for retry.

Supported explicit numerical equalities in mathematical and other written explanations now receive a bounded exact arithmetic check. Supported notation includes literal sums, differences, multiplication, division, brackets, decimals and percentages. Currently asserted exact errors can veto acceptance. Corrections, quoted false examples and stated approximations still require contextual judgement, which can be wrong. Unsupported notation, rounding, source truth and broader mathematics are not certified. Complete written explanations must fit the 6,000-character limit; working supplied through a handwriting activity must fit 2,000 characters. Overlong responses are returned for shortening rather than silently clipped. An unmarked response is not a wrong attempt or proof that its content is incorrect.

Written-answer checks now request structured results to reduce unreadable replies. An unreadable result can still use the next configured service within the same check. The same source and assessment checks still apply. If no usable result is available, the answer stays unmarked and is not counted as wrong. These checks do not guarantee that every judgement is correct. Longer written responses receive the same source and assessment checks in smaller sections, keeping the complete wording and context. Check feedback carefully: correct surrounding statements do not make a reversed relationship or an unsupported claim correct.

Reviewed worked models are prepared in the lesson display format before generation. Their protected text, records and figures still have to match the reviewed source; adding display fields does not permit changing those contents.

The same bounded arithmetic checker now screens lesson tutor replies, including cache hits, before display or narration. Unsafe replies receive a general checking hint; this adds no extra AI request and does not guarantee every tutor error is detected. Help requests also provide contextual calculation flags to make the next checking question more specific, without treating a quoted mistake as an endorsed claim.

For mode and median lessons, generation and review guidance now calls for independent frequency tallies, correct odd/even middle calculations, an explicit convention for tied or non-repeated values, and a distinction between a maximum and a mode. Completed worked methods and results must be visible before different fresh tasks. These instructions help review; they do not guarantee every statistics error is caught.

Choice-feedback checks now flag some explicit English references to displayed option positions. Quote distinctive answer content in feedback so it stays meaningful when choices change order; actual source and sequence references remain valid.

Source-based written explanations now require structured evidence checks before acceptance. Unavailable checks keep the answer unmarked for retry without recording a wrong attempt.

Chemical-equation generation and review now include explicit guidance on per-element counts, unchanged formula subscripts, and ionic versus molecular descriptions. Check the whole draft; these instructions do not guarantee that every scientific error is detected.

Selected place-value explanations now request labelled numbers and exact digit-value working for an additional independent check. Missing or ambiguous evidence stays unmarked; review the displayed format before retrying.

Some explicit English single-answer instructions now block review when a multiple-answer task keys several answers. Grammar guidance also checks intended actors, timeline and valid alternatives; these checks do not guarantee that every grammatical ambiguity is detected.

Generated lesson totals now follow their valid activity estimates automatically. Content review remains required, and invalid timing estimates still need correction.

Authoring checks now report inconsistent total/activity time estimates and explicit completion claims such as “Mastered!” in learner headings. Review guidance also checks equivalent answer alternatives, stated mathematical domains and completed worked teaching before fresh practice. Inspect the whole draft and learner experience after review.

Grade 9 source checks are active for the ten reviewed topics. They verify the exact source edition, bounded targets, current mapping and independent-authoring permissions. Source admission does not approve a generated lesson or certify a completed student audit. Ordinary tables now remain visible across activity types, including multiple-answer questions.

The learner library uses the reviewed current source area when available, so the Grade 7 English reading pilot no longer appears under Indigenous Languages. Repeated area names now share one navigation area, with each lesson retaining its own visible source credit or historical-source warning.

Prepared Grade 8 checks now bind the reviewed source subset and required cell/map models to their exact topics. Source permission and current lesson checks remain separate; supplied worked answers cannot replace fresh learner practice.

Lesson checks now reject malformed case-study observations and conflicting classification IDs before learners open an activity. Correct the reported fields and review the lesson again. The England Year 6 circulation model now supplies its four observation rows in the readable format; its source and assessment limits still apply.

Rejected open responses now use bounded checking guidance from the visible task, rather than freeform negative feedback. The marking verdict remains a separate assessment.

Source-coverage errors now identify the outcome and missing activity references so an author can repair the teaching or learner practice before retrying review.

Mathematical written explanations and all handwriting activities now receive one additional independent check before a proposed correct response earns credit. It checks the whole displayed task and each complete required step, including quantitative working, units and domains where applicable. A correct later calculation does not retract an earlier contradictory claim. Each accepted handwriting alternative must satisfy the whole step by itself; matching one partial example cannot complete a multi-part step. Separate independent requirements into separate steps when authoring. Dependent clauses and comparisons are checked with their subjects and the complete local answer, rather than treating a comma-separated fragment as an independent claim. Complete equivalent reasoning remains valid. A bounded check can reject an explicit hour-hand comparison that contradicts the time stated in the same answer. It does not establish the accuracy or completeness of the rest of the response; the normal independent checks still apply. False comparisons and incomplete working still receive no credit. Uncertain checks keep the response unmarked without a wrong-answer penalty. These model checks can still miss mistakes; review the actual learner work and reported issues.

What’s new - 21 October 2026: Complete working and assertion checks

Section titled “What’s new - 21 October 2026: Complete working and assertion checks”

Proposed correct mathematics explanations and handwriting responses now receive an independent check of every claim and complete task requirement. Review multi-part handwriting steps so each accepted alternative is complete; split independently scored outputs into separate steps. Previously accepted fragments may need correction or completion. An uncertain check keeps the answer unmarked, with the work retained.

What’s new - 21 October 2026: Source-backed writing and unmarked responses

Section titled “What’s new - 21 October 2026: Source-backed writing and unmarked responses”

The selected canonical response tasks can now use their exact preceding reviewed teaching models during checking, while fresh-case records keep priority. Unverifiable responses display an uncertainty message rather than being described as a service outage. They remain unmarked, with the answer kept and no wrong-answer penalty or automatic pass. Other source, review and publication checks still apply.

What’s new - 21 October 2026: Concurrent actions and complete working steps

Section titled “What’s new - 21 October 2026: Concurrent actions and complete working steps”

English inhalation ordering checks now hold lessons that assign a strict order to separate diaphragm and intercostal contractions. Group the simultaneous muscle actions before the resulting volume, pressure and airflow changes. Handwriting guidance also clarifies that each accepted form is a complete alternative for one required step; separately required calculations need separate steps. Re-review affected drafts before publishing. These changes do not alter existing activity controls or certify all answer keys.

Evidence-matching validation now checks that any source body is present in the visible source detail. Conflicting hidden text must be corrected before the lesson passes checks. Source labels and bodies now also receive type checks before a lesson opens; malformed objects or numeric fields produce an issue instead of reaching the activity renderer.

The prepared England Year 6 Animals including humans pilot requires ten activities, two reviewed functional models with local text, and one original linked explanation. Incompatible requests stop before paid generation. Source permissions and current lesson checks still apply; these task checks do not certify practical, medical or written understanding.

Prepared Grade 7 topics now distinguish their reviewed source heading from the existing catalogue label in Studio and the editor. An unavailable source heading stays explicit. Grade 7 authoring and delivery still require exact source bindings, permissions and current lesson checks.

The reviewed Grade 4 subtraction and multiplication lessons, and Grade 5 addition, subtraction and multiplication lessons, now check that the required own-pattern task is present, including three transition equations and its linked practical explanation. Older task reviews need repair and a fresh review before the lesson becomes available again.

CBC Grade 10 now lists distinct 2025 pilot source identities for Core Mathematics, English and Chemistry. Existing school defaults are unchanged; the new options still require their separate source-permission and lesson checks. Pilot lessons remain unavailable until their exact reviewed topic, source permissions and current automatic review qualify. A changed source or required evidence can make an earlier lesson unavailable until valid regenerated or re-reviewed content passes.

The prepared Grade 10 Core Mathematics update requires 12 activities for Real Numbers, Indices and Logarithms, and Quadratic Expressions and Equations. Use compatible formats and preserve the quadratic area image and required learner production. Incompatible requests stop before paid generation; source permissions and current checks still apply.

Reviewed KS3 lessons now keep their shared-stage objectives distinct from ShuleOne year placements. The CBC Grade 5 and Grade 6 pilots each admit only ten reviewed topic subsets when their own source permissions are enabled. Cambridge Stage 6 now has a separate ten-topic original-authoring pilot across Mathematics, English, Science and Computing. Its 26 selected objectives retain explicit coverage limits and require actual learner production where specified.

The reviewed Stage 4 Science object-key lesson now checks its supplied observations, complete routes and answers against one fixed model before publication.

Table checks now hold invalid or unlabelled source tables as drafts so you can repair the labels and cells before learners see them.

Failed generation now retains an editable draft where possible, with all validation reasons visible. A batch reports these as failures so you can repair them before filling the remaining places. For handwritten working, each required step needs a marking guide with actual correct worked responses. A step label alone is insufficient: the validation message identifies missing answer forms so the draft can be repaired. Inspect every requested part and its acceptable equivalent working before publishing. Source-release changes now create a separate current-source artifact when you generate the topic again, preserving the earlier draft’s record. Reviewed-subset lessons now distinguish their specific teaching and practice targets from the full official objective. A published lesson or a full starting-set count does not certify coverage of omitted formats or practical skills.

Generated Grade 4+ lessons now need independent evidence for the exact curriculum level and topic. An older passing AI verdict needs a fresh source-grounded review before republication. Missing source evidence or delivery rights blocks publication automatically; there is no new manual approval step. Fill class to 10 is a coverage aid: a count of ten does not certify curriculum alignment, subject balance or the learner experience.

Grade 4+ generated lessons now publish after their AI quality and narration checks pass. The manual release review screen and its approval step have been removed from the portal. Lessons that fail a check remain visible as drafts in Content.

The class view shows each learner’s position on their quests, what they have completed, and what they are getting wrong.

The last of those is the one to use. It shows a topic several learners are failing on while there is still time to reteach it — which is a different kind of information from an exam result, and arrives about a month earlier.

Look at the class before the individual. Five learners stuck on the same stage is a teaching problem. One learner stuck is a learner problem. They need different responses.

Projects and written work come to the submissions queue.

Open the submission, review the work, award marks and write a comment.

The comment is the teaching. A mark tells a learner where they finished; a comment tells them what to do next. One specific sentence beats three general ones.

Mark promptly. Work marked two weeks later has stopped being about the learning and started being about the grade. Little and often, rather than a weekend of forty.

For coding projects, look at how it was built rather than only whether it runs. A working project assembled by copying is worth less than a partly working one where the learner understood every block.

The coding pathway runs like any other quest, with one difference: learners get stuck in ways that need a different response.

Teach reading the error. Most learners do not read them at all. The error usually says what is wrong and where. Making a class read one aloud before touching the code is the single most useful habit to build.

Do not fix their code. Ask what they expected to happen and what happened instead. The gap between those two answers is where the learning is, and taking the keyboard removes it.

Let them break things. The playground and the sandbox exist for that.

Expect the pace to spread. Coding classes spread further and faster than most subjects. Plan for the fast learners to have somewhere to go — extending their own project usually works better than more exercises.

Set exams from the tutor view: build the paper, set the window, activate it.

For a tab-locked online coding or robotics exam, choose Allow interrupted automatic submissions to continue if learners should be able to recover from a switched-off laptop, closed page or tab-lock submission. You can also change this setting on the exam’s row after creating it.

This reopens the learner’s same attempt and saved answers; it is not a fresh sitting. The original duration keeps running, and the attempt must still be inside both that duration and the exam’s closing time. A learner who deliberately selects Hand in cannot return. While an automatic submission can still be continued, it stays out of the marking queue; it appears for marking when the recovery time ends or when you switch the setting off.

Before a real online exam, run a five-minute practice with the class. Every problem you will have on the day — a device that will not load, a learner who cannot sign in, a corner of the room with no signal — appears in the practice, when it costs nothing.

Marking picks up written answers; the rest is marked automatically. The per-question analysis is worth reading before you hand the results back.

Tutors can now allow an automatically submitted coding or robotics exam to continue after a device or browser interruption. Saved answers return on the same attempt, the original clock does not restart, and deliberate hand-ins remain final.

Some coordinate lessons now show a first-quadrant grid with equal unit scales. In a construction activity, choose the coordinate pair for each named vertex: the grid plots your choices and joins the sides. You can change the choices before checking. A written answer or a description of a drawing does not count as constructing the polygon.

Publication review checks each citation against its actual activity field. An answer’s stored ID and its displayed label are different; invalid review evidence keeps a draft held.

Lesson passages preserve the author’s line breaks and indentation. Read each table row, algorithm step or numbered record on its own line; longer lines wrap on smaller screens. When reviewing a draft, still check that every required row and step is present and readable.

Some afternoon clock-reading activities draw the clock directly from the same source record used for checking. Read both hands and give the requested working and time formats. For these bounded tasks, a fully recognised complete written method can be proved exactly, including every stated calculation, the hand comparison, the displayed clock and elapsed time. Extra claims, incomplete responses and other wording do not receive that certificate; they retain the normal full review. The proof does not establish mastery of a whole topic.

Some fraction and electricity activities now show a diagram that changes as the learner chooses an answer. In a fraction activity, the controls shade equal parts of equal-sized wholes. In a simple series circuit, each control places a standard component symbol in one of four connected positions. Learners can inspect and change their diagrams before checking the answer. These are bounded classroom representations, not physical experiments.

When reviewing a generated lesson, check every required topic against its visible teaching and fresh practice. A passage describing a picture does not replace the picture. A written answer box cannot collect a drawing. Publication checks now hold those mismatches, ambiguous fraction choices and incorrectly keyed equivalent time quantities. If a lesson is held, edit the actual teaching or activity and review it again before publishing.

Learners not attached to a class — self-directed, or from outside the school — appear separately. They are visible if your school has enabled it, and they progress at their own pace without set work.

The portal is at its best when it is doing something the classroom cannot: letting each learner move at their own pace, and telling you what they are getting wrong before an exam does.

It is at its worst when it is used as the whole lesson. Learners working through quests in silence for forty minutes are not learning much that they could not have learned at home.

Use it for practice, for the learners who need to go back, and for the diagnosis. Keep the teaching.

Before publication, lesson checks reject conflicting weekday/date pairs, constant-velocity answers contradicted by their supplied endpoints, and single-choice questions offering interchangeable reciprocal expressions. References to a shown angle or anatomical diagram need an actual supported figure; generic text-clue cards do not count as that figure. These are bounded checks. Still inspect the complete teaching, task, alternatives and feedback against the selected curriculum outcomes. A passing automated review is not a guarantee that the lesson is correct.

For supported exact discrete-value frequency tables, reviews also receive independently calculated totals, cumulative frequencies, median positions, tied modes and mean/range arithmetic. These calculations help reviewers check the supplied records; they do not establish curriculum coverage or certify the whole lesson. Grouped interval tables are not treated as exact observations.

An exactly repeated review receipt can be collapsed only when every required activity is already covered and the repeated record is identical. The original response is kept alongside the verified receipt, and stored evidence checks recompute that relationship. Missing activities, conflicting duplicates, changed findings and failed checks still block publication; this cleanup does not make another AI request or remove a content finding.

Thirteen further Cambridge Stage 6 science topics have reviewed source subsets for model use, enquiry, living processes and ecosystems, material properties, light, circuits, rocks and soils, the Moon, and scientific evidence in context. Each keeps the complete framework group as context and states which targets the digital lesson actually assesses. Source admission is separate from successful generation and publication.

A supported series-circuit interaction does not certify parallel-circuit construction; a bar chart does not certify scatter-graph skills. Supplied Moon records are distinct from a learner’s observations over a real month. Puberty and infection lessons use general educational examples, without asking learners to disclose personal development or medical history. Physical experiments remain unassessed by supplied-record reasoning.

A lesson that asks learners to label a diagram must provide an actual supported visual. Text clue-matching remains available when the task explicitly asks for text clues. Science authoring and review also check that cycles have a stated starting convention, classifications concern the named material, and environmental trade-offs have supplied evidence for both benefits and costs. Every generated lesson still needs its own checks.

Review of a written or classification task now keeps the original finding’s full activity in focus and prevents unrelated choice-key roles from replacing that review. The original finding stays blocking if its evidence cannot be resolved against the correct activity. A dynamic simulation’s chosen inputs also cannot be marked using unresolved variable names in a static handwriting key; use the actual lab readings or a separately stated fresh input. These checks apply wherever the shared engine is used.

Short citation locators are now explicitly labelled when they show only a prefix of a longer field. Reviewers must inspect the full supplied definition before reporting a cut-off prompt; a short citation is not evidence of a learner-visible truncation.

Further England upper key stage 2 source scope

Section titled “Further England upper key stage 2 source scope”

What’s new — 23 October 2026: Selected Year 5 and Year 6 English word-reading and transcription topics, both years’ working-scientifically topics, and Year 5 human development now have separately reviewed source subsets. Lessons still need to pass the current source permission, content, learner interaction and publication checks before appearing for students.

English lessons use contextual word-family reasoning and typed spelling or supplied reference tasks. Typed answers do not certify pronunciation, fluent reading aloud, physical handwriting or use of a writing implement. Scientific enquiry is taught through the relevant year’s substantive science content, with clear distinctions between supplied records and measurements a learner has actually made. Human development examples teach general changes and individual variation without requesting personal health information. These selected lessons do not certify the whole programme or all practical skills.

Contextual word, reference and setup checks

Section titled “Contextual word, reference and setup checks”

What’s new — 23 October 2026: Shared lesson authoring and content review now explicitly check who or what a derived word describes, spelling changes and the complete context. Reference tasks must supply usable entries and meaningful evidence; a study, article or interview label alone does not replace its content. Spelling production and reference checking should use distinct material so copying a supplied completed word is not mistaken for independent spelling. These are review requirements; they do not guarantee every generated claim is correct.

The renderer check also holds requests to label parts of an experiment setup when no actual setup figure is supplied. Authors can provide a supported real visual or rewrite the task explicitly as matching written equipment-function clues. The check applies to generation, editing, review and publication across curricula.

What’s new — 23 October 2026: Selected England Year 4–6 design-and-technology topics have exact reviewed source subsets for user needs and criteria, material selection, product evaluation, reinforcement reasoning and ingredient origins or seasonality. The official DfE source’s public licence is recorded separately from lesson approval. A source release does not publish lessons; each lesson must pass the normal checks before students can see it.

These lessons use complete supplied design briefs, property records and clearly fictional user feedback. Written planning or material selection does not certify physical making, accurate tool use, load testing or cooking. Origin and seasonality tasks do not assess a learner’s nutrition or cooking techniques. Practical and wider programme requirements remain outside these selected lessons.

Earlier worked teaching in written-answer marking

Section titled “Earlier worked teaching in written-answer marking”

What’s new — 23 October 2026: Written explanations in current exact-source lessons now include their actual preceding read-only worked models as teaching context in both marking passes. Earlier inclusion was restricted to protected canonical tasks. The fresh question and its records still govern the fresh case; a worked example does not replace its numbers or observations.

Only preceding case-study passages, data and visible observations are included. Later activities, answer keys, feedback and other learners’ answers are excluded. Existing protected canonical model checks remain in place. A teaching-context change changes the marking cache identity. Marking also explicitly distinguishes correct numerical trends from a reversed causal explanation. These checks support review; passing them alone does not prove that every explanation is correct.

What’s new — 23 October 2026: Missing-figure checks also cover promises to label parts of a product and captions claiming a diagram shows an object. Generic icon or text cards cannot replace that promised picture. Supply a supported actual figure or make the task explicitly about matching written clues.

Publication review now more explicitly keeps a fresh task’s stated dimensions and criteria separate from its worked example. When a finding discusses several answer alternatives, adjudication must address them together under that original finding index. Duplicate or missing indexes still leave the review held; the system does not silently drop a finding or turn key membership into proof of correctness.

What’s new - 23 October 2026: Bounded geography practice

Section titled “What’s new - 23 October 2026: Bounded geography practice”

England primary geography source support now separates selected written coordinate, water-cycle and grid-method practice from regional studies and practical fieldwork. Teachers must still inspect and publish each draft through the normal source, review and media checks. This support does not certify complete geography coverage.

What’s new - 23 October 2026: Visual mode and coordinate boundaries

Section titled “What’s new - 23 October 2026: Visual mode and coordinate boundaries”

Generator guidance now distinguishes plain clue matching from supported rendered figures. A bounded shared check holds literal coordinate classifications whose keyed north/south hemisphere contradicts the supplied latitude, including zero latitude. These checks cover specific formats and do not replace full content review.

What’s new - 23 October 2026: Map and grid-line checks

Section titled “What’s new - 23 October 2026: Map and grid-line checks”

Missing-visual checks also cover explicit map and compass-rose labelling. A bounded glossary check flags definitions that reverse the orientation of easting and northing grid lines, while allowing descriptions of the horizontal or vertical coordinate position.

What’s new - 23 October 2026: Branching water pathways

Section titled “What’s new - 23 October 2026: Branching water pathways”

A bounded check holds generic water-cycle ordering tasks that force runoff and infiltration into one chain without a specified route. A fully supplied particular water-parcel route can still support ordering; naming a starting stage alone does not impose an order on independent branches.

What’s new - 23 October 2026: Written marking response reliability

Section titled “What’s new - 23 October 2026: Written marking response reliability”

Source-grounded written marking uses native strict response shapes on supported OpenAI models, including the separate assertion check. This reduces malformed review responses; it does not establish that an answer is true. The server still checks every required criterion, answer unit and evidence reference. Missing fields, unsupported claims or uncertain checks cannot silently become credit. If a response remains unmarked, its text is retained and it does not count as a wrong answer; retry when the checker is available.

What’s new - 23 October 2026: Additional Grade 5 science source checks

Section titled “What’s new - 23 October 2026: Additional Grade 5 science source checks”

Selected force-and-energy lessons require independently reviewed original-authoring scope and current source permission before generation. An earlier pilot approval does not authorise the newly reviewed topics. Source approvals advance without relabelling historical lesson records. Complete source outcomes are kept separate from the narrower written tasks. A passing written lesson does not replace teacher observation of the practical outcomes.

Open My classes → Class lists and select the school, class and optional stream. Coding admins can choose any coding-programme school; tutors see only their assigned students. An exam is not required.

Print / PDF uses the same class-list report as School management → Registration → Reports → Class List, including the school letterhead, summary and stream sections. Choose Blank columns (PDF) from 0 to 10 for handwritten attendance or marks. Export Excel uses the same formatted management workbook. Each export includes only the roster you are authorised to see, even when your assignment covers one stream.

Use View certificates beside a student to download their certificate, or Class certificates for one printable PDF containing the earned certificates for the selected class and assigned streams. These use the same Coding, Robotics & AI certificate design and award rules as school management.

Year-end certificates become available from September. Awards are calculated from learning progress and marks across terms 1–3: Bronze requires at least 30% completion and a 50 average; Silver 50% and 65; Gold 75% and 75; Platinum 90% and 85. Students who have not earned an award are omitted from the class PDF. If nobody qualifies, the page explains this without downloading an empty file.

In My classes → Class lists, choose the school and use Reporting cohort to switch from Current classes to the cohort before a saved promotion. Then choose the former class and, if needed, its stream. The historical list includes pupils now in Alumni and excludes pupils who were later promoted into that class.

Select Report term and use Report card for one pupil or Class report cards for the whole selected class. Certificate downloads, PDF attendance lists and Excel exports use the same selected cohort. Historical streams still restrict tutor access. Return to Current classes for today’s register.

For example, Lakewood Emerald’s cohort before the 14 October 2026 promotion contains 125 former Grade 9 pupils. Its current Grade 9 register contains the 87 former Grade 8 pupils. Choose the former cohort to download Grade 9 reports and certificates.

Existing exam reports retain their saved exam membership, even after pupils leave. Only a recorded, applied promotion can be selected; reversed promotions are excluded.