Open source is the last honest idea in tech. We build something people can inspect, improve, and adapt. Let developers take it further than one company ever could alone. Linux became the backbone of huge parts of the modern internet that way. Android won by being open enough for hardware makers to run with it. Open source although generous it is really about scale. It is a growth strategy that works because openness has helped technology travel.
A lot of companies want the benefits of being seen as open. They want developer trust, momentum, startups building on top of their tools, researchers doing free ecosystem-building work around their products. But when the scaling happens, a lot of that openness starts to narrow.
Meta leaned hard into the idea that its AI strategy would stand apart because of openness. Llama made Meta look like the company willing to push powerful AI into the open while competitors kept their best models locked behind APIs and subscriptions. It bought Meta credibility with developers and made open source AI part of its brand.
This week’s Muse Spark launch shows how the strategy can change with competitive pressure. Meta’s new flagship model is being rolled out first through its own app and ecosystem, while access for outside partners is limited to a private preview. Meta has not disclosed key technical details like model size, and larger models in the Muse family may come later and could be open-sourced in the future. Muse Spark is not just a chatbot model. Meta is positioning it to power its own app, website, and eventually WhatsApp, Instagram, Facebook, Messenger, and smart glasses.
Open source is powerful, but it is also risky. If you give away too much, someone else can repackage your work and capture the upside. So companies use openness to build momentum, then tighten control around the parts that actually make money, look at Android.
Linux succeeded because it was open enough for an entire ecosystem to form around it. Android succeeded because Google understood that scale mattered more than control and with scale control will come. In AI, companies spending billions on talent, chips, and infrastructure want to be the platform everyone builds on without giving up the best parts of what they built. At least in the U.S. market. There are many open source Chinese and some European open source models. Meta however wants to become the AI standard.
Read my blog about Claude Mythos. This story about open source AI and Claude Mythos makes you really think to what extent open source AI should be pushed.
https://techcrunch.com/2026/04/08/meta-debuts-the-muse-spark-model-in-a-ground-up-overhaul-of-its-ai
https://www.theverge.com/tech/908769/meta-muse-spark-ai-model-launch-rollout
Discover more from Shayan Hosseini
Subscribe to get the latest posts sent to your email.
