Hyper — architecture

Hyper proposes, tests, and iterates on novel ML methods with no human in the inner loop. Given a problem — a formal definition, an evaluation suite, and the current SOTA — it surveys the literature, proposes first-principles alternatives, validates a proof of concept, then implements and evolves the survivors until they beat the best baseline. Before expensive ML experiments it calibrates on lower-stakes domains like board games and combinatorial optimisation, then lifts onto low-compute ML tasks.

Two abstractions hold the loop together. A Problem is the tuple (definition, evaluation suite, current SOTA). An Approach is a research direction defined by a SOTA deficiency, a proposed alternative, and a falsifiable hypothesis that predicts the improvement. Everything below moves Approaches through the loop and pins their implementations to a canonical code library, so fitness differences reflect the strength of the principle rather than implementation noise.

QD Archive quality-diversity population store seeded with SOTA behavior space diversity → Knowledge Graph Literature structure Experiment traces Approach clusters Underexplored regions underexplored Code Republic coding agents · impl. library stable sub-problem modules amortized across experiments open-source commercial API { · · · } Problem definition · eval suite · current SOTA Hypothesis Generator Literature Search survey · extends KG Approach Population diverse candidates Filtering AI debate · one-shot exps Approach SOTA deficiency · alternative · H1 PoC Agent Hypothesis Breakdown dependency graph of verifiable sub-questions Fast Experiments pre-prototype checks rapid falsification CODER Agent repair loop · max_debug_depth debug_prob · Code Republic QD Loop full implementation · eval suite · QD scoring · evolve(poc, population) · ShinkaEvolve problem archive scores H1 disproven informs reads / updates blocks
Figure 1 — Hyper's discovery loop. A Problem enters the Hypothesis Generator, which emits an Approach. The PoC Agent kills weak hypotheses cheaply before any full build; survivors pass to the QD Loop, which implements and evolves them and returns scores to the QD Archive that seeds the next generation. The Knowledge Graph informs proposals and records results; Code Republic supplies the canonical implementation blocks that keep the principle signal above implementation noise.

Sub-systems

Hypothesis Generator

searches research directions

PoC Agent

proves it before building it

QD Loop

implements · scores · evolves

QD Archive

diverse population store

Knowledge Graph

organises what's known

Code Republic

canonical implementation library

CR / Hyper synergy — how Code Republic and Hyper reinforce each other.
Brain vs. LLM — the existence proof for the region of design space Hyper searches.