Three things that happened in AI this week, and what they actually mean for you. Five minute read, or hit play and listen.
This week the tools shifted under everyone's feet again. Nothing you built has to break because of it, if you set it up right.
1. Claude Went Down Again, and That's Actually Useful Information
Anthropic's Claude had another outage this week, hitting errors across multiple models. It's not the first time, and it won't be the last time any AI service does this.
Here's the part that matters for you. If your whole workday depends on one chat window being up, an outage isn't an inconvenience, it's a stop-work order. That's a fragile system, not a real one.
A solid AI setup treats the model like a swappable part. Your prompts, your files, your process, those are yours. The model is just whoever's answering the phone today.
Your move this week: Write down, in one sentence, what you'd do in the next 5 minutes if your main AI tool went dark right now.
2. Your Favorite Model Has an Expiration Date
Moonshot AI is retiring Kimi K2.5 at the end of August. Its replacement, Kimi K3, is a much bigger 2.8 trillion parameter model taking over the lineup.
Models retire. That's just the pattern now, and it applies everywhere, not just Kimi. If you've spent months learning one model's quirks instead of building a workflow that travels, that time can evaporate overnight.
The fix isn't picking the "right" model and hoping it lasts. It's keeping your prompts and process somewhere that isn't locked inside one app, so swapping the engine underneath is a five minute job, not a rebuild.
Your move this week: Copy your three most-used prompts out of the chat app and into a plain text doc you control.
3. Gemini's Having a Loud Week (For Better and Worse)
Google's Gemini 3.5 Pro may be launching any day, Gemini 4 is already quietly in training, there's chatter about a 4x price gap between Claude Sonnet 5 and Gemini Flash, and at least one reviewer this week called several Gemini models flat out bad.
That's a lot of noise pointing in different directions, and it's a good snapshot of the whole industry right now. Everyone's shipping fast, everyone's got an opinion, and the rankings change before you finish reading them.
Chasing the headline of the week is a losing game. The people getting ahead are the ones who test things themselves on their own real task, cheaply, and decide from there.
Your move this week: Open Google AI Studio and run your most common task through two different models back to back, see which one actually wins for you.
The thread that ties it together
Outages, retirements, price wars, hot takes, it's a lot of movement in one week. But none of it should touch your actual system if your prompts, files, and process live outside any single app. The model is the rental car. Your workflow is the license.
If you want the short version of how to build that kind of setup from scratch, grab our free First 7 Days guide. It walks you through owning your own AI system in a week, no chasing headlines required.
Want the practice plan that goes with this? Start with our free First 7 Days with AI guide. One small win a day, exactly where to click, no jargon.
Sources this edition: This edition drew on Anthropic Claude News coverage of the Claude outages, Moonshot AI's announcement on Kimi K2.5 retiring for Kimi K3, and Google Gemini reporting on Gemini 3.5 Pro, Gemini 4 training, the Claude Sonnet 5 vs Gemini Flash price comparison, and Theo's Gemini model ranking.
Want AI actually working for you by next week?
Start with the free guide. One simple step a day, no coding, built for non-technical professionals.
Get the First 7 Days guide