The Weights Do Not Lie: How Rio's 'Sovereign' AI Was Unmasked by a Laptop Audit

When the municipal IT department of Rio de Janeiro, IplanRIO, announced the release of Rio-3.5-Open-397B on Hugging Face, the accompanying press release read like an epochal shift in Latin American technology. Here, supposedly, was a "frontier-class municipal AI model" developed in-house by local public servants, fine-tuned from Alibaba’s Qwen 3.5 base to beat state-of-the-art open models on global benchmarks. Political leaders and municipal pride celebrated a sovereign triumph. There was only one problem: the…

Tech Briefing — Monday, June 15 2026

Open-weight model innovation continues to pick up steam as both MiniMax and Moonshot AI drop highly sparse, huge-scale MoE models targeting multi-modality and long-horizon tool use. However, engineering evaluations warn that massive context windows still suffer from acute mid-document retrieval blind spots, and developers have even flagged local governments repackaging open models as custom builds. Meanwhile, broad global developments highlight rising ebike tensions in Australia, a rejected population cap in…

• 🤖 Models and Launches 4
• 🔥 Hacker News 4
• 🌍 World and Australia 6
• 🔒 Security 1

Corporate Memo Satire

Today's blind comparison was a delight, highlighting how well current models can inhabit complex, deadpan satirical personas on strict constraints. While two candidates stood out with spectacular, laugh-out-loud technical justifications for their 'homegrown' model, we saw one disqualifying style break from a model that insisted on chatting before writing.