Quick follow-up to my Fable shutdown post 👇 The news: Anthropic's Managing Director of International, Chris Ciauri, said at a press conference in Seoul they're "very confident that in the coming days, the models will become available again." Real Anthropic exec, on the record — not a leak. After the US government pulled the most powerful public AI four days post-launch, it's signaling a comeback. The context that makes it credible: Anthropic sent engineers to Washington for talks with Commerce, Dario was at the G7 with Trump pushing AI coordination, and prediction markets put ~58% odds Fable returns before July 1. My take: "Coming days" is doing a lot of work — could be tomorrow, could be two weeks. I read it as "the deal is close," not "set your alarm." The bigger question isn't when — it's what comes back. It got pulled over a cybersecurity jailbreak, so the realistic path isn't "flip it back on." It's "restored with conditions" — extra safeguards, a vetting layer, or narrower access. Possibly a more locked-down model than the one that shipped. For builders and operators: This is my whole thesis playing out live. Model access is now political — a frontier model's availability can hinge on a meeting in Washington. Which is exactly why you build model-agnostic. The devs who got burned wired production straight to Fable on day one. The ones who were fine switched to Opus 4.8 and kept moving. Be the second kind. Fact — Ciauri said "coming days," on the record.Open — the date, and what condition it comes back in. "Too dangerous to release" → "released anyway" → "government pulled it" may be about to hit "it's back, but different." I'll bring you the confirmation when it's real. If Fable 5 comes back more locked-down than launch — does that change whether you'd build on it? 👇

Hey everyone! Mike here and I'm on a journey to learn automations. It's new to me and I'm eager to learn here!

Most people open Claude and just... start typing. Whatever's in their head goes straight into the chat. Sometimes it works. Usually it doesn't. Here's what I do instead: Step 1 — Write the idea in one sentence. Not a prompt yet. Just what I actually want. No formatting, no instructions. Step 2 — Add the role. Who should Claude be when answering this? A copywriter? A strategist? A skeptic? One line. That's it. Step 3 — Add one constraint. What do I not want? Too long, too formal, too generic? Saying what you don't want is often more useful than saying what you do. Step 4 — Read it out loud. If it sounds weird when you say it, it'll produce a weird answer. Focus on the sentence, not the output. That's the whole process. Four steps, five minutes, and the first response is almost always usable. What does your prompting process look like?