Dec 30, 2025

Meta brings in ManusAI as agentic dev tools gain traction

🧩 The Gist

Meta is adding ManusAI to its ranks, signaling continued investment in applied AI talent. On the developer side, new tools focus on running multiple coding agents in parallel and giving assistants longer‑lived context. A widely shared essay argues that AI works best on clean, well‑structured code, reinforcing a push toward best practices. The pattern is clear: orchestration and memory are becoming table stakes for agentic development.

🚀 Key Highlights

• ManusAI announced it is joining Meta via a company blog post surfaced on Hacker News.
• “AI Is Forcing Us To Write Good Code” argues that AI tooling benefits when teams follow best practices, aligning code quality with assistant effectiveness.
• Superset introduced an open source terminal to run 10+ coding agents in parallel on a local machine.
• Superset uses isolated git worktrees, built‑in notification hooks, and a diff viewer to review changes and create PRs quickly.
• Superset is built with Typescript, Electron, xterm.js, and node‑pty, and the team is exploring cloud VM worktrees, context sharing, and a top‑level orchestration agent.
• Ensue Skill adds a shared memory layer to Claude Code that persists context between sessions and supports semantic and temporal search.
• Ensue installs through Claude Code’s plugin marketplace and is positioned as alpha software.

🎯 Strategic Takeaways

• Big tech strategy: Bringing specialized AI teams in‑house remains a priority for platform companies, reinforcing a build plus acquire approach to applied AI capability.
• Agentic workflows: Developers want to spin up many agents at once, isolate them cleanly, and coordinate outputs, which pushes terminals and CLIs to become orchestration surfaces.
• Memory as a feature: Persisting context across sessions is moving from nice‑to‑have to default for coding assistants, improving continuity and reducing re‑prompting.
• Engineering discipline: Cleaner architecture, tests, and consistent structure increase the usefulness of AI pair programmers, tying developer experience to assistant performance.

🧠 Worth Reading

• AI Is Forcing Us To Write Good Code: The core idea is that AI assistants perform better when codebases follow established best practices. The practical takeaway is straightforward, invest in maintainable structure and tests so assistants can navigate projects reliably and produce higher‑quality changes.