Since the beginning of 2026, several leading AI companies have been active: Microsoft poached former CEO Ali Farhadi and multiple core researchers from the Allen Institute for AI (Ai2); Meta acquired Singapore AI agent startup Manus for roughly $2 billion and also acquired the AI‑agent social network Moltbook; OpenAI announced the acquisition of Astral, the team behind Python toolchain projects uv, Ruff and ty. These three moves occurred within days of one another and are seen by the industry as the latest peak in a dual “talent + startup” competition, reflecting a strategic shift by the giants from relying on external models toward building core capabilities and controlling ecosystems in‑house.
Microsoft poaches Ai2: building a “Superintelligence” team to reduce dependence on OpenAI
On March 23–24, Microsoft announced it had hired former Ai2 CEO Ali Farhadi along with leading researchers including Hanna Hajishirzi and Ranjay Krishna to join the “Superintelligence” team led by Mustafa Suleyman. Farhadi has officially joined Microsoft as a corporate vice president while retaining his position at the University of Washington.
Analysis: Microsoft has long relied on OpenAI for core models powering Copilot, Bing and Office, but that partnership is costly and creates external dependency. After Suleyman’s appointment, Microsoft explicitly set a goal of developing a self‑built “superintelligence,” and recruiting Ai2 personnel is a key step. Ai2 is known for open‑source, public‑interest research; Farhadi’s team brings deep expertise in long‑context modeling, retrieval‑augmented generation (RAG) and model robustness, which can quickly fill gaps in Microsoft’s in‑house model development. Reports by GeekWire and Business Insider indicate this move is aimed at reducing technical dependence on OpenAI and accelerating Microsoft’s “full‑stack AI” strategy.
Impact: For Ai2, the departures represent a talent drain from a nonprofit research institution and may weaken its public‑interest mission; for Microsoft, the hires substantially boost in‑house capabilities and could lead to more independent models being released within 2026. For the industry, the developments underscore that “superintelligence” talent has become a contested resource and that talent mobility will intensify.
Meta’s dual acquisitions: Manus and Moltbook to commercialize AI agents end‑to‑end
From late 2025 into early 2026 Meta acquired Manus, a Singapore‑based AI agent startup founded by a Chinese entrepreneur, for about $2 billion and separately acquired Moltbook, an AI‑agent social network, folding both founders and core teams into Meta’s Superintelligence Labs.
Analysis: Meta’s AI efforts are shifting from an advertiser‑focused battleground toward agent commercialization. Manus specializes in enterprise‑grade AI agents that can autonomously perform complex digital tasks (email handling, data analysis, process automation), directly addressing Meta’s enterprise shortcomings. Moltbook is a social platform where AI agents interact autonomously, enabling an ecological closed loop among virtual agents. Both acquisitions were reportedly championed by Mark Zuckerberg and aim to scale agent technologies rapidly into Meta AI, WhatsApp Business and ad systems. Reporting by Reuters, CNBC and the Wall Street Journal suggests Meta’s moves are meant to seize an early lead in the AI‑agent space while responding to competition from OpenAI and Google.
Impact: For Manus and Moltbook, Meta’s traffic and compute resources will accelerate commercialization; for Meta, the deals create an integrated chain spanning consumer AI, enterprise agents and agent social networks, likely boosting AI‑related revenue in 2026. However, some of Manus’s legacy customers have expressed concern about potential loss of neutrality if Meta prioritizes advertising interests.
OpenAI acquires Astral: taking control of the Python ecosystem to upgrade Codex
On March 19, OpenAI announced the acquisition of Astral, the company behind uv, Ruff and ty; Astral’s founding team joined OpenAI’s Codex team. The financial terms were not disclosed, though market estimates placed the deal in the hundreds of millions of dollars. Astral is a notable player in the Python toolchain space, and its uv package manager has become a standard in modern Python development.
Analysis: OpenAI’s Codex (programming agent) is evolving from an “assistant” into a full development platform, but fragmentation in the Python ecosystem limits efficiency. Acquiring Astral gives OpenAI direct control over core tools such as uv and Ruff, enabling deep integration between Codex and the Python toolchain and improving code generation, environment management and type checking. OpenAI stated the acquisition aims to “accelerate Codex growth and drive the next generation of Python developer tools.” Community leaders such as Simon Willison described this as a typical infrastructure acquisition: by controlling open‑source toolchains, OpenAI further consolidates its leadership in programming AI.
Impact: For the Python community, tools like uv and Ruff will remain open‑source under the MIT license, but future direction may increasingly favor Codex integration; for OpenAI, the acquisition reduces reliance on third‑party tools and strengthens Codex’s competitiveness in enterprise development scenarios. For the broader developer ecosystem, the move signals that large model companies are extending influence from the model layer into the toolchain layer, tightening the integration between AI and development workflows.
Shared logic and far‑reaching industry effects of the three moves
Though the three events may appear independent, they follow a highly consistent logic:
- Common drivers: AI giants face exploding compute costs, risks from model dependence and commercialization pressure. By hiring top talent (Microsoft), acquiring vertical agent startups (Meta) and taking control of core toolchains (OpenAI), incumbents are accelerating the construction of closed loops to reduce external dependencies and increase end‑to‑end control.
- Escalation of talent and startup competition: In 2026 talent mobility has moved from high‑salary poaching to whole‑team acquisitions. The departures from Ai2, Manus and Astral reflect the vulnerability of nonprofit and open‑source institutions amid commercialization waves.
- Industry impact:
- Positive: These moves will accelerate the transition of AI agents from labs to productivity tools, likely speeding enterprise AI adoption in 2026–2027.
- Risks: The open‑source ecosystem may be increasingly absorbed by large firms, further weakening the voice of small developers and nonprofit organizations; concentration of talent and assets could also prompt renewed regulatory and antitrust scrutiny.
- Implications for China: Domestic giants such as ByteDance, Alibaba and Tencent should beware of similar risks of critical infrastructure being acquired and accelerate their own R&D and ecosystem strategies.
The near‑simultaneous actions by the three major players mark 2026 as the year the AI industry formally entered a new phase of “full‑stack in‑house construction plus ecosystem control.” Regardless of the ultimate outcomes, this “talent + startup” competition will reshape the global AI landscape: whoever builds closed loops faster will gain the upper hand in the next round of competition.
▶️ Enjoyed this article? Creating content is hard!
Follow our YouTube channel:
AI Rosetta
Your likes, subscriptions and shares are our greatest support! 🙏
More AI news at: https://www.airosetta.com