Claude Fable 5's free deadline, a smarter default model, and a $150M bet on people

Anthropic released Claude Fable 5, a new top tier of model it's calling its most capable yet. The headline for you isn't the benchmark scores — it's the calendar. Fable 5 is included free on Pro, Max, Team, and Enterprise plans only through June 22. After June 23 it moves to usage credits until Anthropic has the capacity to make it standard. So if there's a big, knotty build you've been putting off — the one that needs the model to hold a lot in its head at once — this is the window to lean on the strongest engine for free.

One caveat worth taking seriously: early users agree Fable 5 is excellent on long, multi-step jobs and overkill on simple ones. Its deeper thinking is always on, which makes it slower and pricier per task. The practical move is to reach for it when a build genuinely stalls, and stick with a lighter, cheaper model for everyday drafting and small tools. You don't need the most powerful model for every task — you need the right one for the task in front of you.

The default Claude got better, too. Opus 4.8 is now the standard model on most paid plans, and it's noticeably stronger at real work than the version before it. If you built a tool a few months ago and parked it, it's worth opening it back up — the same instructions may produce visibly better results now, with no extra effort from you.

Agents that improve between runs. Anthropic introduced a feature it calls "Dreaming": a scheduled process where an agent reviews its past sessions, spots patterns, and tidies up what it remembers, so it gets better over time instead of starting cold every run. This is the plain-language version of "set it up once and it keeps sharpening" — the exact kind of leverage we teach, no code required.

Recurring tools that run themselves. Claude's Managed Agents can now run on a schedule — a tool fires every morning, does its job, and stops, with no separate scheduler for you to build or host. This one lives on the more technical Claude Platform, so it's for the advanced end of the room, but it's a clear signal: "a tool that runs on its own" keeps getting easier to set up.

If you've built anything that calls Claude automatically on a subscription plan (rather than through the developer API), your billing changes starting June 15: that kind of automated use moves to a separate monthly credit pool, split out from your normal chat usage. Worth a two-minute check on how your tools sign in, so a scheduled job doesn't quietly stop running next week. (Exact limits are still firming up — confirm against Anthropic's own docs before you change anything.)

Anthropic committed an initial $150M to [Claude Corps](https://www.anthropic.com/news/claude-corps), a fellowship that trains 1,000 people to use Claude well and places them full-time, in person, at nonprofits for a year — fellows earn $85,000 plus benefits, and host organizations get a grant and free credits. Applications are open now and close July 17 for the first cohort. We're flagging it because it's our entire thesis at national scale: the limit was never the model, it's people who know how to put it to work. A food bank forecasting its distribution, small teams building tools that used to be out of reach — that's not a pitch, that's what happens when ordinary people get their hands on this.

And a stat that's been making the rounds: Anthropic says Claude now writes more than 80% of its own code, up from under 10% in February 2025. Make of it what you will, but it's a vivid marker of how fast hands-off building has come in about a year. (source)