AI news digest: what changed this week and what to try
AI moves so fast that keeping up feels like a second job, and half the headlines are demos that never ship. This prompt has your todo.is agent follow official announcements, research papers and trusted newsletters, then send a digest that says what actually launched, what is only announced, and one thing worth trying in your own work.
The prompt
- Send me an AI news digest every [SCHEDULE] on [WHERE TO SEND IT]. I use AI for [HOW YOU USE AI]. Cover new models and major updates from the big AI labs, new features in the AI tools I use ([AI TOOLS I USE]), notable open-source releases, important research in plain words, and policy or regulation news. Use official announcements and research pages as primary sources and link them. Label each item "available now", "announced" or "research only". Add one "try this" idea for my work, and one claim going around that is overhyped or unconfirmed. Keep it to [LENGTH].
What to change
- [SCHEDULE]: e.g. "Friday at 16:00" for a weekly digest, or "weekday at 8:00".
- [WHERE TO SEND IT]: Email, Slack, Telegram, WhatsApp or Discord.
- [HOW YOU USE AI]: e.g. "marketing copy and images for a small agency" or "coding and data analysis".
- [AI TOOLS I USE]: e.g. "ChatGPT, Claude, Midjourney, GitHub Copilot".
- [LENGTH]: e.g. "a 5-minute read" or "10 items max".
Example result
- AI digest · week of Oct 6
- For: marketing copy and images at a small agency
- Models and major updates
- • A leading AI lab released a new mid-size model with a longer context window and lower price than its previous model. Available now in the app and API. (official announcement)
- • An image model update improves text inside images, a common weak spot for posters and ads. Available now for paid plans. (release notes)
- • A video model with sound was shown in a demo reel. Announced, no release date. (blog post)
- Tools you use
- • Your writing assistant added project folders that remember brand guidelines across chats. Available now.
- • Your coding assistant now explains pull requests in plain language. Available now.
- Open source
- • A new open-weight image model is small enough to run on a good laptop. License allows commercial use with conditions; read it before using it for clients. (model page)
- Research in plain words
- • Paper: a university team found that asking a model to list its assumptions before answering reduced factual errors in their tests. Small study, interesting idea. (paper link)
- Policy
- • New guidance in the EU on labeling AI-generated images in ads. Worth reading if you run campaigns there. (official page)
- Try this
- Write a one-page brand guide (tone, banned words, 3 sample paragraphs) and paste it at the start of every copy request. It is the cheapest quality upgrade you can make.
- Overhyped this week
- A viral post says a new model "replaces designers". The demo used heavy manual editing. No evidence it works out of the box.
How to do it with todo.is
- Copy the prompt and say how you use AI and which tools you rely on.
- Send it to your agent on email or Slack, or paste it on the Today screen in todo.is.
- Your agent reads official sources and sends the digest on your schedule, each item labeled and linked.
- Ask follow-ups like "explain item 3 in more detail" or "test the try-this idea on my last brief".
Tips for a better result
- The "available now" vs "announced" label saves you from chasing features you cannot use yet.
- Weekly is often better than daily for AI news. Many stories get corrected or expanded within days.
- Name your tools. Feature updates in the apps you already pay for are usually the most useful news.
- Ask for primary sources first. Announcements and papers are more reliable than reposts.
AI news digest: FAQ
- How can I keep up with AI news without spending hours? Pick a weekly digest from primary sources, focused on the tools and tasks you care about. Your agent can do the reading and send a short summary with links.
- What is the difference between an announcement and a release? An announcement shows something is coming, often with a demo. A release means you can actually use it now. Your digest labels each item so you know which is which.
- Can it explain research papers simply? Yes. Your agent reads the paper page and explains the idea, method and limits in plain words. Ask for more detail on any item.
- Does todo.is use these AI models itself? todo.is runs your own AI agent, which can search, read and create files for you. The digest covers the wider AI world, not just todo.is.
JavaScript is required to use the todo.is app.