Claude Fable 5 changed how we work on the Claude Code team day to day. We used to verify that Claude did the work right. Now we verify that it's doing the right work. Here’s the 3 biggest changes:

I’m Putting it to the test.

Why can't I see it in the desktop app yet?! I have no patience lol.

Anthropic right now

It's refusing to scan our apps for security vulnerabilities sadly

Please do text instead of video. Easier to consume on the go

moving form "checking if the work is right" to "checking if it's doing the right work". the ideas guy era is upon us

Everyone's talking about Claude Fable 5's benchmarks. I'm more interested in the workflows that won't exist a year from now because of it.

Fable as planner and code reviewer is about all we can afford.

The time has come…

Bırak kolpayı limitleri sıfırla amk

It doesn't even work. It reroutes all tasks. An excuse to get users conversation history If they - while annoyed submit feedback. The sabotage of anyone doing AI-research is a huge red flag. In general, the model is too expensive and in no way useful as is. Nor do I feel comfortable using it. This is giving me post 9/11 security theater vibes all over. Hype up fear of model capability to justify backend restrictions that violate user rights. I'm not impressed.

How do you use Claude for video editing?

Let me guess: loops?

Looks like I am not testing Fable much 😭

The real change is that verification moves upstream. You now need tighter intent specs and executable contracts before the session starts or the autonomy just executes the wrong problem faster.

@BlnaryMlke It’s just too expensive haha Burned through my Max + 100€ of extra usage in like 1h and never finished the work… 🫠

@cursor_ai bench data here is impressive for score but even more for the drop in cost per intelligence. Fable5 Med $8.27/69.8% vs Opus 4.7 Max $11.02/64.8% Fable5 Low $5.70/64.2% vs Opus 4.8 Max $7.59/63.8% The trend about the price per intelligence dropping is continuing.

Can we use claude for video editing?

Had it scan a project for bugs and it found a minor security thing. Instantly shut down... and now that session is toast because the security thing is in the context. I read the docs as saying I'd be downgraded to Opus?

I don't know what is real or fake @trq212 anymore after the office episode

"we used to verify it did the work right. now we verify it's doing the right work." that one sentence describes the entire shift in how we build.

@mattpocockuk they basically use your grill-me skill, it is becoming more and more important as models get smarter!

Cool, now we need a e8b and 30b mini Claude’s to run locally in conjunction with the main cloud model so it can offload minor remedial tasks while leveraging the users hardware to accelerate the user facing side

Wonder where they got that idea from cc @mattpocockuk

"If there's something you think LLMs couldn't do, give it a chance" Such good advice @trq212 !

'verify it did the work right' vs 'verify it's doing the right work.' one is a QA problem. the other is a judgment problem.

Testing it now on the same real task I gave Opus earlier today. Early impression: I’m correcting less and trusting it more. Let’s see if that holds through the final result.

Insane stuff, congrats guys. Looking forward to testing Fable 5.

you stop checking if it's right, you start checking if it's even trying to do the thing you asked

brb taking out a mortgage to afford a 4th Max plan

“Fable can run hours at a time” Yeah for those that can causally dump $10k lol

Now Claude is verifying our work, not the other way around. We’re officially just ideas guys now 😂