This week’s spotlight is on Latam-GPT, a first-of-its-kind open-source LLM developed across 20 Latin American countries and Spain. Built with a massive 50B-parameter architecture and trained on regional data—including indigenous languages—it’s less about hype and more about technological sovereignty and community empowerment. A big swing toward AI that reflects cultural realities, not just Silicon Valley defaults.

Cohere pushes multilingual accessibility with Aya 23, Microsoft reports AI assistants are now indispensable for 80% of devs, and Meta quietly looks to Google and OpenAI as it preps Llama 5. On the consumer side, Google’s cheekily named Nano Banana turns heads with pro-grade image edits inside Gemini, while Microsoft unveils its first in-house AI models, signaling less dependence on OpenAI and more vertical integration.

And… Anthropic…

An open‑source LLM built for regional relevance — the first of its kind.

A large language model initiative led by CENIA (Chile’s National Center for AI), developed collaboratively across 20 Latin American countries and Spain. The 50B‑parameter model is trained on an 8+ TB regional corpus.

It represents a breakthrough in culturally adapted, sovereign AI solutions tailored to Latin American needs—including inclusion of indigenous languages like Mapuche, Rapanui, and Guaraní.

Latam‑GPT aims not just for AI capability, but for regional empowerment and educational accessibility by reducing reliance on global tech giants.

“Prioritizes technological sovereignty, regional collaboration, and educational development.”

The first version is slated for release in 2025—so its impact depends on timely deployment and sustained support.

Cohere’s non-profit arm released Aya 23, an open‐weights LLM supporting nearly two dozen languages. This marks a significant step for multilingual accessibility, expanding open model availability beyond English-centric datasets.

Microsoft research shows 75% of developers regularly use AI coding assistants, with 90% reporting higher productivity. Notably, 80% said they’d be upset if AI tools were suddenly removed—signaling their deep integration into daily workflows.

Meta is exploring partnerships with Google and OpenAI to supplement its own AI models as it develops Llama 5. Early signals suggest Meta may blend external tools to bridge capability gaps in the short term while iterating internally.

Microsoft debuted two home-grown AI models: MAI-Voice-1, a GPU-efficient speech generator producing one minute of audio in under a second, and MAI-1-preview, a large language model trained on 15,000 Nvidia H100 GPUs. Both are in active testing, signaling Microsoft’s move toward independence from OpenAI.

Nano Banana, the codename for Gemini Flash 2.5 Image, launched with powerful editing tools: seamless multi-step edits, consistent characters across frames, and natural background adjustments. Available in the Gemini app, users call it “state-of-the-art” for pro-level edits in seconds.

Another week of separating AI signal from noise. If we saved you from a demo that would've crashed prod, we've done our job.

📧 Got a story? Reply with your AI tool wins, fails, or war crimes. Best stories get featured (with credit).

📤 Share the skepticism: Forward to an engineer who needs saved from the hype. They'll thank you.

✍️ Who's behind this? The Augment Code team—we build AI agents that ship real code. Started this newsletter because we're tired of the BS too.

🚀 Try Augment: Ready for AI that gets your whole codebase?