# AI News Digest - 2026-05-04

1. MIT researchers published a study that provided a mechanistic explanation for why large language model performance scaled reliably with model size, attributing the effect to a phenomenon called superposition.

2. Microsoft inserted a "Co-Authored-by Copilot" line into Git commits created in Visual Studio Code, including for developers who had disabled Copilot's AI features.

3. Xiaomi released the open-weight MiMo-V2.5-Pro, which the company reported nearly matched Anthropic's Claude Opus 4.6 on coding benchmarks while using 40–60% fewer tokens and supported hours-long autonomous coding on single tasks.

4. A US government agency reported that China was eight months behind in the AI race according to its benchmark, while independent data did not corroborate that finding.

5. Researchers released a 100-scenario benchmark that evaluated frontier language models on everyday ethical dilemmas and found that leading models diverged in their moral judgments.

# References

1. [https://the-decoder.com/mit-study-explains-why-scaling-language-models-works-so-reliably/](https://the-decoder.com/mit-study-explains-why-scaling-language-models-works-so-reliably/)

[MIT study explains why scaling language models works so reliably](https://the-decoder.com/mit-study-explains-why-scaling-language-models-works-so-reliably/)

1. [https://the-decoder.com/co-pilot-becomes-a-co-author-in-vs-code-without-being-asked/](https://the-decoder.com/co-pilot-becomes-a-co-author-in-vs-code-without-being-asked/)

[Microsoft caught sneaking "Co-Authored-by Copilot" into VS Code commits - even with AI off](https://the-decoder.com/co-pilot-becomes-a-co-author-in-vs-code-without-being-asked/)

1. [https://the-decoder.com/xiaomis-open-weight-mimo-v2-5-pro-takes-aim-at-claude-opus-with-hours-long-autonomous-coding/](https://the-decoder.com/xiaomis-open-weight-mimo-v2-5-pro-takes-aim-at-claude-opus-with-hours-long-autonomous-coding/)

[Xiaomi's open-weight MiMo-V2.5-Pro takes aim at Claude Opus with hours-long autonomous coding](https://the-decoder.com/xiaomis-open-weight-mimo-v2-5-pro-takes-aim-at-claude-opus-with-hours-long-autonomous-coding/)

1. [https://the-decoder.com/china-is-falling-behind-in-the-ai-race-according-to-a-us-government-benchmark/](https://the-decoder.com/china-is-falling-behind-in-the-ai-race-according-to-a-us-government-benchmark/)

[China is falling behind in the AI race, according to a US government benchmark](https://the-decoder.com/china-is-falling-behind-in-the-ai-race-according-to-a-us-government-benchmark/)

1. [https://the-decoder.com/same-prompt-different-morals-how-frontier-ai-models-diverge-on-ethical-dilemmas/](https://the-decoder.com/same-prompt-different-morals-how-frontier-ai-models-diverge-on-ethical-dilemmas/)

[Same prompt, different morals: how frontier AI models diverge on ethical dilemmas](https://the-decoder.com/same-prompt-different-morals-how-frontier-ai-models-diverge-on-ethical-dilemmas/)

For the site tree, see the [root Markdown](https://ixtj.dev/.md).