Fable Returns With Limits artwork

Fable Returns With Limits

The Daily AI Show

July 1, 2026

The hosts opened on Q3, Canada Day, and the expected return of Fable with usage limits and possible code-related restrictions. They compared Sonnet 5, Opus, Fable, Codex, Claude Code, Hermes, compound engineering, and GStack as different ways to plan, build, and route AI work.
Speakers: Brian, Beth, Gareth
**Brian** (0:00)
Hey, what's going on, everybody? Welcome to The Daily AI Show. Today is July 1st, which officially means it's Q3.
I work in sales, so I am always very much aware of when the new Q is coming out or starting. So I hope everybody is getting some well-earned rest in the sales community, who is probably trying to push pipeline and get things closed yesterday. It's also Canada Day, so happy Canada Day to all our folks up North, which currently includes Beth. You know, so happy Canada Day, I guess, Beth, since you're sitting up there.

**Beth** (0:36)
My first Canada Day in Canada, so yay, happy Canada Day to all.

**Brian** (0:42)
So I had to go look it up because I didn't know what Canada Day was, but cool background to it. You know, it came out and started as Dominion Day in the mid-1860s when some of the, I'm going to mess it up now, some of the provinces were independent but still part of the British Empire. But then when Canada officially became, had its constitution, everything they renamed to Canada Day. I'm sure I'm screwing that up. I shouldn't even start it. But anyway, happy Canada Day to all our folks listening to us up in Canada. So we hope you have a great day off and you get some fireworks and all the good stuff there.
So it's Beth and myself today. I'm Brian. I think we'll see if Andy can pop. Normally, we would have Jyunmi on a Wednesday, but he had some things going on, he couldn't make it today. So, Speedy Recovery to Jyunmi, and hopefully he's back next week. We'll get to see him, because we always like our AI and science stories when he brings those.
But Beth, we were just talking before we went live, a couple different things going on over the last couple of days, but you were talking about Fable, so why don't we kick it off there, and sort of like what you know at this point, or what you're expecting, because I'll be honest, I've been a little bit out of the loop on that side, so I feel like maybe you can. To speed.

**Beth** (1:59)
So my understanding is that Fable is back. It's been released, the Commerce Secretary released the embargo or whatever the description, the formal description is, its government name. And then we, okay. Hey, Andy.
So that release, Anthropic announced that it would be back today. Now, if you look at X, I just looked, Matt Berman was like, Fable? So today is sort of broad now that we're in the today. And so I don't see it back. I don't have access to it. It should be back today. There's also limitations and I haven't seen the official announcement, but my understanding is that between now and July 7th, if you have an anthropic subscription, you can use half your allotted tokens with Fable, but you can't go over that. And there will of course be some refusals, maybe some lots of refusals for code. Also, I saw something that said code would be downgraded to 4.8. It would go down to Opus when it's writing things.
And that sort of gets into what Brian and I were talking about before we went live, which is, where do you use your smartest AI?

**Brian** (3:28)
Yeah.
Yeah. And you were saying, we were saying we're probably the opposite. And I won't die on my heels whether I'm right or wrong. But I was saying, oh, so another new story was that Sonnet 5 had released yesterday. And which is great because when you look at Sonnet 5, it seems to be very agent forward. When you looked at the benchmarks, it was in some cases not cybersecurity in other areas, but in some cases very comparable to Opus 4.8, but at a much cheaper cost. And this is something we see over and over again, where the lesser models catch up and then the frontiers move again. And so I played with it a little bit yesterday. I don't really have anything huge to report back on other than for what I was doing in Claude Code yesterday and working on a project. I did not notice a difference, but it's always really like that's so subjective because like, what was I working on and would I notice a difference? It's not like I was running them in parallel, you know? So how would I really know if I was noticing a difference? It's just it's based on Spidey Sense, I guess, which isn't super scientific at all. But you know, that's about what I have. I don't have benchmarks and stuff. So I typically that's how I will, I will, you know, kind of gauge things, which is again, just opinion more than anything else. There's very little fact in there. But but in that conversation, Beth, you and I were talking about, I said, oh, I thought maybe for this particular project, this is a side project for my family.

57 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000775070975