AI Updates

What changed, who should care, and what to do next.

I scan the noise so you can focus on what matters.

Featured update16 July 2026Test carefully

ChatGPT adds a clearer reasoning-effort choice

The useful choice is how much effort a task deserves, rather than memorising every model name.

What changed

OpenAI began rolling GPT-5.6 Sol and Sol Pro into ChatGPT reasoning modes while the everyday default remains a faster option.

Who should care

People comparing difficult options. People using ChatGPT for longer analysis.

Try this today

Run one real task at a fast and a deeper setting, then compare usefulness, elapsed time and correction effort.

What changed

OpenAI introduced Work for longer tasks across selected files and connected apps, with progress, redirection and approval points.

Who should care

People preparing recurring reports. People creating presentations from several files.

Try this today

Test one reviewable deliverable with approved files and require a checkpoint before the full output is produced.

What changed

Moonshot released hosted K3 access across Kimi products and its API while fuller technical material and downloadable weights were still pending.

Who should care

People working across very large file sets. Technical teams designing matched long-horizon tests.

Try this today

Keep it on a shortlist and use a controlled file or research test before changing an existing workflow.

The human take

AI can draft, summarise, and suggest. But you bring the context, the nuance, and the values. Use AI to amplify your thinking, not replace it.