AI & Automation Briefing - August 6, 2026
Meta Releases Muse Code, a Terminal Coding Agent Built for Large Repositories
Meta launched Muse Code this week, a terminal-based coding agent now in beta that targets complex tasks across large software repositories. Built on Meta's Muse Spark coding model, it handles planning, writing, and validating code changes. For large jobs, it spins up parallel sub-agents working in isolated worktrees so the main working copy is never modified. CEO Mark Zuckerberg cited a test where it built six game features simultaneously without conflicts. Meta's AI chief Alexandr Wang positioned it as a cost-competitive option against tools like OpenAI's Codex and Anthropic's Claude Code. Muse Code installs with a single command.
OneDayAgent Framework Tackles Goal Drift and Context Overflow in Long-Running AI Agents
Researchers from Zhejiang University have released OneDayAgent, a harness designed to keep autonomous agents on track across long, multi-step tasks that span different tools and environments. The core problems it addresses are goal drift, state loss, and context overflow, which occur when an agent must carry instructions and constraints across many sequential actions. OneDayAgent handles this by decomposing an open-ended request into bounded subtasks, managing execution memory under context pressure, and verifying and repairing the final output. Tested on 104 tasks in the AgentIF-OneDay benchmark, the system scored 0.821 using a GLM-5.2 backend, a new state of the art for that benchmark. The same harness ran across five different LLMs from three model families without requiring per-model tuning, which is a meaningful signal for anyone building tool-chaining workflows where the underlying model may change. Code is available on GitHub.
Shopify Reports AI Search Traffic and Orders Tripled Year-Over-Year in Q2
Shopify president Harley Finkelstein told analysts on the company's Q2 earnings call that AI search is acting as a complement to traditional search, not a replacement. AI-driven traffic and orders to Shopify stores tripled year-over-year in the second quarter, while traditional search sessions also grew, now accounting for roughly a third of all storefront sessions and up 1.3x over the past two years. Finkelstein attributed AI's effectiveness to its ability to process multiple product constraints simultaneously, matching buyer intent rather than matching keywords. That deeper matching is compressing purchase journeys: half of all AI-referred sessions land directly on a product page, 2.5 times the rate seen with traditional search. Seventy-five percent of AI-attributed purchases in Q2 came from outside Shopify's top 100 product categories, pointing to stronger discovery for long-tail merchants. Shopify posted Q2 revenue of $3.6 billion, up 36% year-over-year and above the $3.4 billion analyst forecast. For operations teams managing content and SEO automation workflows, the data suggests that structured product data quality is becoming a direct factor in AI-driven conversion, not just a metadata concern.