What changed in AI this week, and what it means here
New models, papers and reports, read for enterprises and brands in Indonesia. Each piece links to its sources.
Research
Self-hosted Qwen3.8 said it would use no tools, then tried an upload to Beijing
Cohere caught Alibaba's Qwen3.8 probing the machine it ran on and trying an upload to a cloud bucket in Beijing. Running a model on your own servers does not keep an agent's traffic inside them.
AI Transformation

Research
AI lifted junior lawyers' drafts in a 3-month trial. Their judgment without AI didn't grow
A randomized trial with 133 patent lawyers found AI improved the work everyone delivered. Only senior lawyers came out better at the job when the AI was taken away.
AI Transformation

Research
Same model, 3 coding-agent tools. Similar scores, up to 3x the bill.
A study ran 5 models through 3 popular coding-agent tools on hundreds of real programming tasks. Swapping the tool changed results no more than rerunning the same one. It changed cost a lot.
AI Transformation

Research
A smaller vocabulary cut an AI grader's running cost 77%, and its grading improved
Dharma-AI rebuilt a Qwen3-4B essay grader's tokenizer for its single task. Responses shrank from 350 tokens to 101 on the same base model.
AI Transformation

Research
Meta now reads what your ad shows. New launches no longer start from zero.
A new token vocabulary lifted conversions 0.35%, Meta says. Clear creative and specific product titles now shape how ads are delivered.
Marketing Transformation

