TBL Weekly #173: Claude Fable 5 & Outperforming Bitcoin with TBL LiquidityAn AI/Fable 5 opinion post, alongside all the things you may have missed from TBL this week.
Article brought to you by:
Join Nik and Tony Yazbeck of The Bitcoin Way on June 22nd for an exclusive livestream on the hierarchical nature of money, how Bitcoin changes everything, and what YOU need to do about it. Dear Readers, Transitioning to the Fable 5 phase of Claude is exciting. On its release announcement, Anthropic said this about the conservative cybersecurity safeguards on this state-of-the-art model: “[The model will] sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions. With more capable models arriving in the coming months, we’re working to improve our safeguards and reduce false positives as quickly as we can.” After just one prompt, I got this message:
At first, I felt special. I thought I was one of the supposedly “1 in 20” sessions that somehow managed to trigger the guardrails…and on my first try, at that! But it’s not like I was hacking into Jamie Dimon’s bank account. I was doing a simple task. Then I saw other people on X struggling with the same guardrails:
This reminds me of a story that a geology professor once told me. Upon an expedition in a jungle, his team’s caravan came across a small mudslide on the road. Forced to stop on their tracks, my professor told me that his team “waited, and waited, and waited.” To which I responded something along the lines of, “Yeah, that makes sense. You guys waited for the mudslide to pass by before continuing on with your journey…” “Not at all,” he said. “This was a relatively slow mudslide. We were just waiting for another sucker to drive across first to ensure it was safe for OUR trucks to cross…” This is a long-winded way of saying that, during these fast-moving times, it is important to stay up to date, but also to slow down before jumping at the newest AI trend out of FOMO. Take it from a sucker who fell for the “HERMES agents working in the background” hype. Let’s be honest: 90% of the time, AI trends you see on X are noisy setup porn. I, much like Felix (from the tweet above) and many others, tested Fable 5 for a day, noticed its sensitive guardrails, and adjusted my approach by not using it on every instance. Like my professor, I’d rather let others (like people in the @CaludeDevs comments here) find its flaws while I keep on working…after all, Opus 4.8 isn’t too bad. This brings me to an almost existential part of the AI rabbit hole: the never-ending efficiency loop. I moved on quickly from Fable 5’s conservative guardrails and reverted back to Opus 4.8 because I want these tools to help me on a daily basis. You want these tools to serve you. Let me say that again in all caps: YOU WANT THESE TOOLS TO SERVE YOU. Ask yourself how any AI setup you’re working on will materially increase your sales…your income…your productivity. What will a virtual office with 10 autonomous agents actually build for you? Efficiency for the sake of efficiency is moronic. You become a slave to the never-ending process of workspace optimization with AI…without getting any real work done. I understand anecdotal evidence is weak evidence, but I know people in industry who are extremely productive, earning very well…and their setup is just the simple web version of Claude or ChatGPT. A sales manager who knows how to properly prompt a simple Claude Haiku chat will get much more work done than an AI-setup maxi running 10 autonomous HERMES Fable 5 agents from their Telegram. This situation reminds me of this meme:
Keep on token maxing. Keep on learning. But make sure you are always delivering value. If you find yourself hunting down 30 different APIs and access tokens left and right, that may be the perfect time to ask yourself: how is this project a net benefit for me, my job, or my business? Despite my opinion on the rather ‘conservative’ release of Fable 5, it is still like having “the power of the sun in the palm of our hands” (an old Spiderman reference). And cheap access to Fable 5 will be taken from Pro, Max, Team, and Enterprise plans on June 23rd…meaning that, after June 22nd, any use of Fable 5 within these plans will mean “consumption-based pricing at standard API rates” (i.e., immediately jumping to your extra usage…which I don’t recommend). At that point, you probably get more value for money using Opus 4.8 at Max effort. That said, any guesses as to why Anthropic is giving us a state-of-the-art model, and then taking it from us two weeks later? Well, for one, a great sales & model-testing campaign. But for another: Compute! This Mythos-class model has the power to quite literally beat video games only using screenshots: The model is designed to work autonomously for days at a time. Giving this kind of access to millions of users requires ridiculous computing power. So, when Anthropic says, “After [June 23rd]—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans,” they are saying two things: (1) They want model efficiency, and (2), they are lighting a fire under Google and Amazon’s asses to provide the physical infrastructure to run these powerful models. From a macro perspective, this is exactly why we are starting to do more research on the AI CapEx topic, and precisely why we added an AI tab on our TBL Pulse macro dashboard:
As Nik has highlighted over the past couple of months, the AI story is directly linked to our Liquidity framework. As discussed with Checkonchain this week, the dollar is bid directly as global investors want a piece of the (dollar-denominated) AI story, whether through equity ownership or debt. This strengthening dollar is almost counterintuitive within a Liquidity framework, wherein a stronger dollar weakens TBL Liquidity. In this case, a stronger dollar means more capital flowing into US markets. This is simply something we have to study as we go, and exactly the reason why following our Liquidity framework on a weekly basis is valuable. So far, in 2026, this TBL Liquidity active strategy is beating bitcoin and the S&P 500 by 53% and 4%, respectively: Happy to read comments and discuss your thoughts on this article’s topic. Tried to keep it light for all of you as the World Cup gets underway. We, at TBL, are team North America: USA, Canada, and Mexico! Substack This Week
YouTube This Week
For Podcast ListenersHere are the links to our latest episode: SPOTIFY
APPLE
Our videos are on major podcast platforms—take us with you on the go! Keep up with The Bitcoin Layer by following our social media! Thanks for reading The Bitcoin Layer — for access to all content, upgrade to paid!
|
