Technology

Self-Improving Harnesses Push ARC-AGI to 95.5%

YC Paper Club put harnesses center stage: Prime Agent, OpenJarvis and QM show how self-improving scaffolding beats fixed prompts.

StartupHub.ai Staff

By StartupHub.ai Staff, reviewed by Daniel Singer, Editor

3 min read

Contents(3)

Prime Agent hit 95.5% on ARC-AGI using self-improving harnesses on Claude Opus, up from 30% on the private set without scaffolding.

The demo at YC Paper Club showed three live systems, Prime Agent, Open Jarvis and YC's QM, all rewriting their own prompts and code. They run as persistent local processes with file, REPL and sub-agent access, so the real risk is local session access, not remote exploit.

The full discussion can be found on YC's YouTube channel.

Self-Improving Harnesses, Local Personal AI And YC's Agent For Work | YC Paper Club - YC
Self-Improving Harnesses, Local Personal AI And YC's Agent For Work | YC Paper Club, from YC

What is AGI, and what does ARC-AGI measure?

AGI, or artificial general intelligence, is AI that can handle a wide range of tasks at roughly human level, instead of one narrow task. ARC Prize defines AGI more narrowly as a system that can match the learning efficiency of humans.

ARC-AGI, the Abstraction and Reasoning Corpus for Artificial General Intelligence, was introduced by François Chollet in a 2019 paper. It measures fluid intelligence by asking a system to learn a rule from a few examples and apply it to a new task. The tasks are easy for people and hard for AI. A score such as Prime Agent's 95.5% is a result on that test, and a high ARC-AGI score does not by itself show that a system is AGI. Sam Altman has called GPT-6 Astra an early step toward AGI.

How self-improving harnesses actually work

The model on its own is a Turing machine, tokens in, tokens out. The harness wraps it in a von Neumann layer: L1 weights, L2 active context and REPL, L3 files and sub-agents, all of which it can garbage-collect and rewrite. DSPy evolves the system prompt through genetic search over small training sets. Darwin Godel Machines evolve the harness code itself and archive winners to breed stronger scaffolds.

Why this matters and what it still doesn't fix

For builders, the payoff is immediate. An 18% framework gain can decide whether Arcade AI ships, and Open Jarvis claims 800x lower cost by keeping inference on an Mac Mini-class device. Qwen 3.8 27B now matches Claude 4.6 Opus from August 2025, shrinking the cloud lead that local harnesses rely on. Nothing here is patched by default. QM centralizes chats in Postgres and gates writes with human review, but auto-approval tends to drift and weak permissions leak secrets, so isolate stores and audit every rewrite.

Self-editing scaffolds work. They just only work inside the boundary you define.

Sources

TopicsY CombinatorPrime AgentOpenJarvisQMDSPyARC-AGIVoyagerToolformer

StartupHub.ai StaffDaniel SingerWritten by StartupHub.ai Staff, Editorial teamReviewed by Daniel Singer, EditorAbout

StartupHub.ai Staff, Editorial team

The staff writers of StartupHub.ai, ranging from investment analysts to avid AI tool users, early adopters and critical enthusiasts. Backgrounds span engineering, business and the arts. We hold every piece to rigorous standards of research and review.

Daniel Singer, Editor

Daniel Singer is an AI expert who has been writing about, covering and building AI technology since 2018.

More from Daniel Singer

Get the free daily brief every morning

Every morning the day's AI news, new AI startups and new funding rounds. Free and takes a few minutes to read, unsubscribe at any time.

Join 64,000 other subscribers.

Test your Agent Readiness now

Paste your website. How well ChatGPT, Claude and the other agents can read and use a site, in about a minute.

The videos behind the news, one flip at a time

The feed mixes the videos from our articles with startup demos and you flip through them like TV channels for free with no signup.