Gal Vered

Founder @checksumai

[Gal Vered @g_vered]

@realAkashAnand research

[Gal Vered @g_vered]

@marckohlbrugge @johancutych Also say it'll fallback to Opus without mentioning whether it'll be transparent and whether you'd be charged opus or fable rates.

[Gal Vered @g_vered]

@marckohlbrugge Yeah I feel like they constantly need to clarify

[Gal Vered @g_vered]

Seems like it'll be transparent. If it clearly shows in the API and claude code, than I think it's reasonable.

[Gal Vered @g_vered]

This is such a dark pattern.

If Fable can't do what I asked of it, reject the request. Don't route it to another model.

[Anthropic @AnthropicAI]

Claude Fable 5 will be available again globally tomorrow.

After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests.

We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort.

Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research.

Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again.

Read our full blog: https://t.co/VHyum831ri

[Gal Vered @g_vered]

@DanielLockyer @ivanjurashere Anthropic comms is so bad. I feel like they constantly need to clarify

[Gal Vered @g_vered]

@0xidanlevin Why illegal? maybe against ToS

[Gal Vered @g_vered]

@mitchellh It's really fun watching you build this with such care. The small details that meke great products!

[Gal Vered @g_vered]

@JackWoth98 @GoogleAIStudio Google AI Studio is really good

[Gal Vered @g_vered]

@picocreator @FeatherlessAI @aiDotEngineer 10 out 10 10 booth!

[Gal Vered @g_vered]

@zachlloydtweets I use Warp daily as a terminal. Time to check out the agents I guess

[Gal Vered @g_vered]

@GergelyOrosz Yes and what features have they shipped during this time that you actually care about? for me zero

[Gal Vered @g_vered]

@mike64_t I don't think I ever saw an AI written test legitimately fail because of a bug

[Gal Vered @g_vered]

@delba_oliveira What is this view though?

[Gal Vered @g_vered]

That's the main problem with the AI hype:

I am a daily user of Spotify. Love this app. But, they have not shipped a single thing I care about in the last 5 years.

It's already great! but I just need it to play songs. So how are the 4500 deploys translate to my life being better?

[ClaudeDevs @ClaudeDevs]

Boris sat down with Spotify VP of Engineering Niklas Gustavsson.

Spotify ships 4,500 production deploys a day, and 73% of PRs are now AI-assisted.

[Gal Vered @g_vered]

@divvsaxena Those are different models so it makes sense.y question is about same model, different harness

[Gal Vered @g_vered]

Is there any alpha in different harnesses? I feel like they all perform pretty much the same

[Gal Vered @g_vered]

@divvsaxena Do you really get better performance using opencode vs codex vs pi vs Claude code assuming same model and skills and tools

[Gal Vered @g_vered]

Making inference 3x faster will provide just as significant progress as making them 3x smarter

[Gal Vered @g_vered]

@dillon_mulroy Yes!

[Gal Vered @g_vered]

@ryanbrewer I have with the official chrome extension. Couldn't do anything beyond basic actions. What do you use it for that it's successful?