Automated AI News Brief: GPT-5.6, Copilot, and Inference Services

Introduction

This brief was created from data fetched by Horizon, then selected and rewritten by Codex. Horizon only handled collection.

1. GPT-5.6 Sol and Luna updates in ChatGPT

OpenAI improved GPT-5.6 Sol and expanded GPT-5.6 Luna access for free users. When model availability changes, use fixed tasks to compare quality, speed, and cost instead of switching settings on model names alone.

Source: OpenAI: GPT-5.6 update

2. ChatGPT moves from asking to doing work

OpenAI shared How the world is putting ChatGPT to work. Verifiable outputs, data permissions, and human confirmation matter more than handing an entire workflow to a model.

Source: OpenAI: ChatGPT at work

3. Kimi K3 arrives in GitHub Copilot

GitHub announced Kimi K3 in Copilot. As model options grow, evaluate them by task type and keep human code review for security-sensitive changes.

Source: GitHub: Kimi K3 in Copilot

4. Baseten joins Hugging Face Inference Providers

Hugging Face announced Baseten as an Inference Provider. A unified inference entry point helps switching, but latency, failures, data boundaries, and cost still need monitoring.

Source: Hugging Face: Baseten

5. Datasette continues updating

Simon Willison released Datasette 1.0a38 and 0.65.3. For agents reading data, clear schemas, limited queries, and separated read/write access are safer than broad database permissions.

Sources: Datasette 1.0a38; Datasette 0.65.3

Today's Notes

AI is moving further into daily tooling. The durable difference will come from model choice, data access, and observability keeping up.