Root-cause groups

Table options / naming parity drift (10 tasks)

1. What the model wrote:

   const damageEvent = table(
     { name: 'damageEvent', public: true, event: true },
   

   const player = table(
     { name: 'player', public: true },
   

   const membership = table(
     {
       name: 'membership',
       public: true,
       indexes: [
   

   const nextPosition = table(
     { name: 'nextPosition', public: true },
   

   const categoryStats = table(
     { name: 'categoryStats' },
   

   const result = table(
     { name: 'result', public: true },
   

   const announcement = table(
     { name: 'announcement', public: true },
   

   const author = table(
     { name: 'author', public: true },
   

   const player = table(
     { name: 'player', public: true },
   

   const primitive = table(
     { name: 'primitive', public: true },
   

2. What was expected:

   const damageEvent = table({
     name: 'damage_event',
     public: true,
     event: true,
   }, {
   

   const player = table({
     name: 'player',
   }, {
   

   const membership = table(
     {
       name: 'membership',
       indexes: [
   

   const nextPosition = table(
     {
       name: 'next_position',
   

   const categoryStats = table({
     name: 'category_stats',
   }, {
   

   const result = table({
     name: 'result',
   }, {
   

   const announcement = table({
     name: 'announcement',
     public: true,
   }, {
   

   const author = table({
     name: 'author',
   }, {
   

   const player = table({
     name: 'player',
   }, {
   

   const primitive = table(
     {
       name: 'primitive',
   

3. What the error says:
   The scorer consistently reported schema_parity: describe failed / spacetime sql failed / spacetime call failed.

4. Why this happened:
   The model repeatedly used the wrong table option shape and, in several cases, invented extra options like public: true or event: true. It also failed to normalize names to the expected snake_case schema names in some tasks, likely overfitting to inferred runtime semantics instead of matching the benchmark’s exact schema definition.

5. Affected tasks:
   t_024_event_table, t_025_optional_fields, t_014_elementary_columns, t_015_product_type_columns, t_029_filter_and_aggregate, t_012_spacetime_product_type, t_031_unique_constraint, t_019_many_to_many, t_020_ecs, t_022_view_basic, t_033_sort_and_limit, t_028_cascade_delete

6. AI guidelines gap:
   The guidelines appear to under-specify the exact table({...}, {...}) option object format and when public should be omitted. They also do not emphasize strict name normalization rules such as camelCase code identifiers versus snake_case table names.

Reducer callback signature/style mismatch (6 tasks)

1. What the model wrote:

   export const seed = spacetimedb.reducer((ctx) => {
   

   export const activeAnnouncements = spacetimedb.anonymousView(
     { name: 'active_announcements', public: true },
     t.array(announcement.rowType),
     (ctx) => [...ctx.db.announcement.active.filter(true)]
   );
   

   export const build_leaderboard = spacetimedb.reducer(
     { limit: t.u32() },
     (ctx, { limit }) => {
       const players = [...ctx.db.player.iter()];
   

   export const step = spacetimedb.reducer((ctx) => {
   

   export const seed = spacetimedb.reducer((ctx) => {
   

2. What was expected:

   export const seed = spacetimedb.reducer(ctx => {
   

   (ctx) => {
     return Array.from(ctx.db.announcement.active.filter(true));
   }
   

   export const build_leaderboard = spacetimedb.reducer(
     { limit: t.u32() },
     (ctx, { limit }) => {
       const players = Array.from(ctx.db.player.iter());
   

   export const step = spacetimedb.reducer(ctx => {
   

   export const seed = spacetimedb.reducer(ctx => {
   

3. What the error says:
   The failures still surfaced as schema/parity errors, indicating the generated API shape did not match exactly.

4. Why this happened:
   The model introduced unnecessary block formatting and, in one case, used an array spread where the benchmark expected Array.from(...) and explicit return. This suggests it was approximating JavaScript patterns rather than reproducing the benchmark’s required idioms.

5. Affected tasks:
   t_014_elementary_columns, t_022_view_basic, t_033_sort_and_limit, t_020_ecs, t_019_many_to_many, t_015_product_type_columns

6. AI guidelines gap:
   The guidelines likely do not stress that reducer/view callback syntax must match the benchmark’s exact style, including concise arrow formatting and explicit return in anonymous views.

Type/name normalization errors in tables and indexes (4 tasks)

1. What the model wrote:

   const damageEvent = table(
     { name: 'damageEvent', public: true, event: true },
   

   const categoryStats = table(
     { name: 'categoryStats' },
   

   const nextPosition = table(
     { name: 'nextPosition', public: true },
   

   const result = table(
     { name: 'result', public: true },
   

   const spacetimedb = schema({ entity, position, velocity, nextPosition });
   

2. What was expected:

   { name: 'damage_event', ... }
   

   { name: 'category_stats', ... }
   

   { name: 'next_position', ... }
   

   { name: 'result', ...