Discussion across the Atmosphere
Claude Fable is relentlessly proactive
Simon Willison’s Weblog · simonwillison.net
After two days of experience with Claude Fable 5 I think the best way to describe it is relentlessly proactive. It knows a whole lot of tricks and it will …
Read on simonwillison.net52 Bluesky posts · 1 Semble save
Discussion
-
Josh Collinsworth @collinsworth.dev
This is a great example of what I'm talking about. AI enthusiasts would likely get lost in everything the agent is doing; "look how productive it is! It's inventing whole debugging flows!" But step back, and you see it burned $12 in tokens on a 2-line CSS fix.
-
tj mahr 🦃 @tjmahr.com
@jessicahullman.bsky.social in the comments: "I’ve had similar experiences – as soon as model fitting runs into a problem, coding agents start devising their own idiosyncratic solutions" oh no. this plus [relentless proactivity] is a nightmare
-
Marcus @marcusreed00.bsky.social
This AI's proactivity is wild. Smart, but terrifying if it gets subverted. Reminds me why sandboxing is crucial. #saas #startup
-
Bruce Sterling @bruces @bruces.mastodon.social.ap.brid.gy
*WTF
-
Camille Roux @camilleroux.com
Claude Fable 5 laissé seul sur un bug CSS : il ouvre Safari, écrit ses propres pages de test, injecte du JS dans les templates, monte un serveur HTTP pour récupérer des mesures via CORS. Zéro instruction pour tout ça.
-
a security treat @turbocri.me
and then there's stuff like this this would be a tall tale, if it were about any other model. idk, i keep my agents on a tight leash, so maybe i just never get a chance to watch that kind of thing
-
MudBlossom @lomai.bsky.social
Have you seen this? It’s quite…something.
-
Manton Reece @manton.org
Wild example of Fable being relentlessly proactive on Simon Willison’s blog. I remember the first time I saw AI writing a little throw-away program to analyze something. Taken too far, it almost has the feel of paperclip maximizing… Achieve the goal at all costs.
Saved in
-
🬕🬢▌🬜🬎🬓🬕🬢▌🬴🬰🬃🬰🬴🬐 @nandi.uk