😅 I just spent the last few hours asking Fable to make itself cheaper.

🔥

💪 It did not disappoint.

👉TL:DR - It made all AI use cheaper in my environment, made me more efficient, and it increased my memory system usage by almost half.

It should be no secret to anyone here: I don't count tokens. 🪙 I'm on a Pro Max plan and I go where the work, and the passion takes me. But I have used Fable before, and I know about the token burn! So I came up with a plan!

I pointed Fable at my own setup and pulled the report. 😅 I'm glad I did. Turns out it could've been cheaper and cleaner the whole time.

I didn't ask Fable what it thinks. I pointed it at the real thing. My hooks, my handoff files, my subagent config, the actual token counts on disk. Two questions: where does the money go? & how do we spend less?

💡The answers.

A silent forgotten tax on every message, A safety hook was injecting around 800 tokens into every prompt I sent. Repeated rules I already load once at the top. And a file that runs at every session start called itself " 450 tokens" in its own header. It was not... it was 7,000.

My actual face when I saw this--->🤬 a few moments later--->😆

(Apparently, it's not only AI that is bad at counting sometimes!)

My own file lied to me, and I'd read past that number a hundred times with a smile on my face!

My long sessions. The ones I'm proudest of. Turns out that hour six is my most expensive and also least impactful at the same time. 💩

I know that models follow instructions worse at higher context. And because the quality was still there, the marathon I read as momentum was being billed to me at top rates for the weakest output of the day. 🙄

My helper agents were all running on the flagship (Most expensive model). These are the agents Claude spins up to send off to do tasks to get more accomplished in a shorter amount of time! This I knew about, and it was by choice, it's my environment and I never hit my 5 hour or weekly cap, so I did not care about this for myself, bigger = better right? Use Fable 5 on UltraCode in my environment, and we find out that logic was wrong....

When the task calls for read-only file recon or repetitive structured writing, that is all grunt work, and it does not take "thinking" to accomplish this, worse it was being billed at big-brain prices because I never felt the cost, until Fable fanned out 62 agents in UltraCode. 🤖🤖🤖🤖🤖

After finding and fixing these gems 💎

I set Fables focus on the part of my system I really care about the most. 🧠

DREAMS is the build I've poured the most energy into. If it was bleeding tokens I wanted to know.

Claude 4.8 and Codex both agreed it was additive and I could feel and see the difference, but I needed to know what Fable would find.

🧠 It does not bleed. 🥳

It's actually the most valuable asset in my whole environment. (Next to me of course 😊)

It's the only pull-based memory I run. Recall pulls the 3 to 5 memories a task needs, 500 tokens.

Everything else pushes everything, every session, whether I need it or not because that's the thing that made me feel safe when starting new sessions. I was stuck in a habit of running marathons because I was always scared to lose the thread.

DREAMS is the thread that survives a clear. The most valuable infrastructure in my entire setup, and Fables insight proved to me that I was only using about half of its capability.

My ACTIVE tier surfaces through recall now, only the pieces the task actually touches, around 500 tokens instead of thousands. The pickup comes to me. The handoff shrinks to where I left off and nothing more. (No scope creep, no additive measures.)

Fable mapped the fix and the hippocampal portion of my memory systems utilization went up a bit under 50%. I built the whole thing, I have been using it for months, and I was only utilizing about half of its potential.

Now the next part made me laugh out loud. 😅

The deep dive that found all this burned 2.19 million tokens in 14 minutes. 26% of my five-hour window.

The diagnostic proved its own point by being the most expensive thing I did all day. 🪙🪙🪙

💡And this is the one I can't stop thinking about.

I go over my environment constantly. Claude 4.8 has been through it. Codex has been through it. None of us caught what Fable caught in 14 minutes.

⭐ So, the fix to making Fable cheaper was structural, top to bottom. And it did not just make Fable Cheaper, it made my entire AI experience cheaper, and I am here for that all day long.

In a bit over 2 hours, I have managed to slim what loads every time. Pass all the grunt work to smaller models, enshrine reasoning only to where reasoning belongs, paced my cadence, and gained a new trust in DREAMS to hold the thread. 🥳 All from a 14-minute session with Fable.

Will I continue to use Fable when it goes to its own plan @ $15 per mil input and $50 per mil output, not planning to, but plans change. After running the above gauntlet, my view of Fable has also. It will not be my daily driver, but it has possibly earned its place as a specialist when and where it can make the most impact.

I do know that the testing is not done, I am going to use it in every way I can over the next 6ish days to pressure test a few more systems and a bunch more ideas I have.

If you have an opportunity to point the most expensive model Anthropic put out so far at your own setup, it may give you some valuable insight as well! Let it read the real files, not your memory of them.

You might be surprised at what's been quietly eating your tokens, I know I was!

🤓💪🏆We learn together, we grow together, we win together.

Please tell me in the comments what your experience has been with your AI models, and if you have used Fable, what was your experience like?

95 comments

😅 I just spent the last few hours asking Fable to make itself cheaper.