Pipeline

From an app's question to a sealed block

Apps ask questions in plain words. Harnesses batch them 500 at a time, merge the duplicates, pack, prove, put them to a jury, seal the rulings into a block anchored to Solana devnet, and send each app its answers.

Blocks

#10 rulings · 0 batches89d789eaad…devnet slot 520,108,578 #20 rulings · 0 batches25a8a734fc…devnet slot 520,108,584

1Apps submit10 appsPlain-word ASC calls, each with pinned context.

2Queue0 waitingWaiting for the next batch.

3Batch500 / 500Fills to 500 requests.

4Pack321 → 3Identical questions merged; the rest packed into prompts.

5Provenot yetOne proof of inference per packed prompt.modeled

6Servenot yetPrompts, context and proof hashes go to the jury.

7Jurynot yetFive seats, four families, commit then reveal.

8Seal and returnnot yetSealed into a block; answers back to each app.

batch-8812 · opened 1:30 · run time 2:00

batch-8812

500 requests from 10 apps collapse to 321 questions, packed into 3 prompts. Each packed prompt is put to one jury; every app that asked gets the answer.

Packed prompt	Lead question	Asked	Answer
1 of 3	Did the Seattle Marathon take place on its scheduled date?	×50	pending
2 of 3	Does the supplier certificate cover the delivered product category?	×50	pending

Proof hashes: 7c621a0c5376 · ec504683821c · d6d2a516c387. In v0.0.1-alpha the proofs are modeled: the hashes stand in for proofs of inference, which exist today for one open model.