AI Coding Live News

Continual learning without weight fine-tuning represents a frontier in LLM agent research. When an agent acts purely by writing and executing Python scripts inside an execution sandbox, every thought, hypothesis, and plan leaves a verifiable trace on disk. In a newly released study (arXiv:2610.11450), researchers subject coding agents to ARC-AGI-3—a suite of instruction-free abstract visual reasoning environments where each sequential level introduces novel causal dynamics. Developing a fine-grained measurement protocol, the authors trace thoughts across file artifacts across seven full evaluation runs spanning three foundation backbones from two major model families. The empirical findings reveal that scripts written for one level are almost never invoked in subsequent levels (only 33 of 630 cross-task references survive), as agents deliberately abandon 74% of scripts at task boundaries to purge level-specific overfitted state while rewriting core rules. Notes are maintained in an append-only fashion, with conflicting claims arbitrated against raw execution logs to achieve selective rather than catastrophic forgetting. Crucially, the study identifies carrying hard-coded variables across task boundaries as the single most devastating failure mode of autonomous agents.

Endowing artificial agents with general intelligence requires inferring environment dynamics and refining world models continuously in the wild. Fine-tuning models directly is prohibitively expensive and precipitates catastrophic forgetting. Researchers from University College London (UCL) and Huawei Noah's Ark Lab UK present Memento 3 (arXiv:2610.11794), a framework enabling frozen LLM agents to achieve model-based Recursive Self-Improvement (RSI) purely via external memory. The agent maintains a natural-language rulebook representing its evolving hypotheses of environment physics, which is compiled into executable code for fast simulation and action planning. Governed by a continuous observation-reflection-revision-compilation-verification loop, the system leverages prediction errors to refine rules. Edits are merged only after passing LLM semantic alignment audits and deterministic cell-exact replay verification. On the rigorous ARC-AGI-3 abstraction benchmark, Memento 3 solves 100% of levels across all 25 public environments, scoring a perfect 100.0 Relative Human Action Efficiency (RHAE) while utilizing only 44% of human action counts. In Atari Pong, a synthesized feedback controller shuts out opponents 21:0 across three evaluation seeds with zero test-time LLM inference.

OpenAI launched GPT-6 Astra, calling it the world's most intelligent and aligned model, able to perform nearly any computer task for the user. It sets new state of the art on FrontierMath Tier 4, ARC-AGI 3, and TerminalBench-4.0, with major gains in software engineering, cybersecurity, science, and professional workflows. It is now live for Pro, Enterprise, Business, and API customers, with Plus rollout underway.

ARC Prize measured OpenAI GPT-6 Astra on ARC-AGI-3 Semi-Private: about 62.7% for ~$26k with the Standard harness and 99.9% for ~$19k with a Provider Adapter harness that preserves opaque reasoning state. Astra beat the human action-efficiency baseline on 96% of levels and built custom algebraic notes and tools.

OpenAI says GPT-6 Astra is state-of-the-art on Agents’ Last Exam, AutomationBench, and ScreenSpot Pro—benchmarks for cross-profession computer workflows and screen grounding—adding computer-use evidence beyond the already announced TerminalBench / ARC-AGI / FrontierMath results.

OpenAI says GPT-6 Astra is state-of-the-art on FrontierMath Tier 4, ARC-AGI 3, and TerminalBench-4.0, and also leads on Terminal-Bench Science 0.1 and HealthBench Pro—highlighting agentic terminal and scientific discovery strength.