This weekly update brings major news about AI coding agents that can write programs by themselves.

Replit, a coding platform, launched Agent 3 this week. This new AI agent can work on its own for up to 200 minutes, compared to just 2 minutes for the first version. Agent 3 can test and fix its own code while building apps with very little help from humans. Replit grew from $2.8 million to $150 million in revenue in less than a year, showing how popular these tools have become.

However, new research shows that AI coding tools might not be as helpful as people think. A study found that experienced developers were actually 19% slower when using AI tools, even though they felt like they were working faster. This is called the "productivity paradox" - the tools feel good to use but don't always make work faster.

Security problems are also growing with AI coding tools. Research shows that AI-generated code has 322% more security risks and gets approved 4 times faster than normal code. This means dangerous code might slip into important programs without proper checking.

Microsoft made a big change by adding Anthropic's Claude AI to its Office programs. This ends Microsoft's exclusive partnership with OpenAI and gives users more AI options for coding tasks.