Prime Intellect released Prime Agent under MIT on August 6, 2026. The self-refining harness runs on a persistent IPython kernel and beats human expert baselines on coding and research benchmarks.

The data is in: AI-assisted developers introduce security flaws at 10x the rate of traditional coders, and technical debt is up 30-41% after AI tool adoption. Here is what the research shows.

Meta launched Muse Code on August 5, a beta AI coding agent powered by the closed Muse Spark 1.2 model. Here is what the pricing, benchmarks, and data trade-offs actually mean for your team.

CapCut PC now integrates Seedance 2.5 for AI video generation, prompt-based scene creation, and timeline editing. Here is what the workflow covers.

Meta launched Muse Code in public preview at $1.25 per million input tokens, with a contributor tier priced 90% lower. Here is what small teams need to know before using it.

TencentDB Agent Memory v2.0 turns conversations, docs, and code into four governed memory assets. MIT licensed, Docker self-hosted, integrates with Claude Code.

Databricks slashed per-task AI coding costs by as much as 90% in some scenarios while growing developer adoption, using four stacked cost control methods.

GitHub's Octoverse report shows one new developer joining every second, 80% using Copilot in week one, and only 30% of AI code suggestions getting accepted.

A browser game with 40,000+ runs shows humans approve roughly one in three malicious AI coding agent requests. npm run analyze fooled 65% of players.

Meta launched Muse Code, a terminal-based AI coding agent for large codebases. Here's what it does, who it competes with, and what else happened in AI this week.

Meta's new terminal-based coding agent runs persistent background sub-agents, logs every action for crash recovery, and is available now via the Meta Model API.

Mark Zuckerberg announced Muse Code, Meta's first AI coding agent, now in beta. It writes code and validates results, putting Meta in direct competition with OpenAI and Anthropic.