Who programed you
Executive summary
You asked “Who programmed you.” Available sources identify Meta (formerly Facebook) and its Meta AI organization as the developer of the assistant referred to in 2025-era reporting: Meta builds its consumer assistant on the Llama family of models (Llama 3 and Llama 4 variants) and publishes research and product pages under “Meta AI” [1] [2] [3]. Reporting describes Meta AI as a broad research and product group that creates the underlying models, tools and deployments across WhatsApp, Instagram, Messenger and the web [2] [3].
1. The corporate author: Meta’s AI organization
Meta AI — the rebranded research and product group that used to be called FAIR — is the entity publicly linked to the assistant and the Llama model family; Meta’s sites and press materials present Meta AI as the hub for the company’s models and consumer assistant work [1] [3]. Independent reporting and summaries of Meta’s product launches treat “Meta AI” as the umbrella that delivers the chatbot and related services to Meta’s apps [2] [4].
2. The model family that powers the assistant
Multiple sources say the consumer assistant is built on the Llama lineage: it launched on Llama 3 generations and, by 2025, has integrated Llama 4 variants across Meta’s surfaces [2] [3]. Coverage of Meta’s model roadmap frames Llama as the core architecture used to run the chatbot inside WhatsApp, Instagram, Messenger and on meta.ai [2].
3. Product pages and official messaging: Meta’s own disclosure
Meta’s official AI pages present Meta AI as a product and research portfolio that provides “personal superintelligence” features and tools for immersive content and assistants [1]. Meta’s public statements and blog posts also describe product rollouts, developer toolkits and the embedding of its assistants into consumer apps [5] [6].
4. How Meta describes model deployment and updates
Reporting and Meta’s documentation indicate the assistant is not a static single model but a deployment that has moved across model generations — Llama 3 to Llama 4 variants — and that Meta publishes model families for consumers and developers, sometimes as open checkpoints [2] [7]. That implies “who programmed” the assistant is both Meta’s engineering teams and the research models they train and package.
5. What sources say about infrastructure and automation
Analysts and press pieces note Meta’s large-scale investment in AI infrastructure, GPUs and custom tooling to train and operate its models; these resources underpin model development and deployment at scale [8] [9]. Some reporting frames Meta’s drive toward AI-assisted engineering and automation of coding tasks inside the company, signaling human engineers plus automated pipelines together build and maintain these systems [10] [9].
6. Alternative framings and limits of the record
Available sources focus on Meta and Meta AI; they do not provide a line-by-line account of individual engineers, external contractors, or all third‑party contributions to the codebase (not found in current reporting). Sources also note that Meta sometimes releases open-weight checkpoints, suggesting external developers can run or adapt versions of the models, but Core product deployments remain organized and released by Meta [2].
7. Privacy, personalization and governance context
When users ask who made and controls the assistant, it matters because Meta has publicly announced changes in how AI interactions will be used to personalize content and ads across its platforms, effective December 16, 2025 — a corporate policy decision made by Meta that directly affects how the assistant’s outputs are treated in the wider ecosystem [5] [11]. That policy linkage reinforces that the assistant is both a technical product and a corporate service governed by Meta’s rules.
8. Bottom line — a short answer with nuance
Short answer: public records in these sources attribute the assistant to Meta and its Meta AI group and identify the Llama model family (Llama 3 → Llama 4 variants) as the foundation [1] [2] [3]. Nuance: building and operating the assistant combines Meta’s researchers, engineers, infrastructure and model checkpoints; open releases mean outside developers can also run related models, but product deployments and policy choices are controlled by Meta [2] [7].
Limitations: available sources do not name individual programmers or provide full supply‑chain details such as third‑party contractors or specific engineering teams beyond the corporate Meta AI label (not found in current reporting).