Do this on your last day with Fable

You can extract Fable 5’s intelligence out of the model before it’s gone… and i’m going to teach you the 5 workflows that do it, with every prompt ready to paste

because tomorrow Fable 5 leaves your subscription

it moves to pay-per-token credits, and if you’re on a normal Claude plan, that arithmetic means one thing: it’s gone for good

so today is the last flat-rate day for the smartest model available, and the way it’s being spent is mostly wrong

the lists going around say build a website, ship a few demo apps, generate a month of content

every one of those fails the only test that should rank your day

the test is below, then the five moves that pass it

[VISUAL: hero, the irreversibility filter as a decision tree]

if you want the full training on getting real business output from models like this, and turning that output into income, that’s what the real time AI ops community is for: weeklyaiops.com

the one test you need to run

one question sorts everything you could do today: can a cheaper model redo this tomorrow?

a website, a demo app, a batch of posts… Opus, the everyday Claude model that stays in your plan, rebuilds any of those next week for nothing

spending the final hours of a frontier model on work a mid-tier model handles is like hiring a surgeon to take blood pressure lol

what a cheaper model can’t redo tomorrow is anything that requires fable-level judgment to create but only ordinary intelligence to use

• a standard written down

• a roadmap already reasoned through

• a knowledge vault already distilled

• a skill that fires on its own

those keep their full value after the model behind them is out of reach

the pattern is old: the most copied training dataset of the llama era was built by pulling 52,000 answers out of a frontier model and training a small open one on them, total generation cost under $500

that frontier model got retired… the teacher is dead, everything trained on it still runs

so the strategy for today is extraction, not conversation

make fable write down everything it knows about your business before it leaves

if you’re starting from zero with this model, the full course covers the setup groundwork:

Machina@EXM7777·Jul 2 ArticleHow to master Fable 5 (Full Course)Anthropic not only just released the best model ever built, they introduced us to a whole new world of possibilities… Fable 5 is a category of its own right now - until GPT 5.6 Sol launches, there’s…2270534245K

now the five moves

1. plant fable’s judgment in your workspace

the highest-value artifact a frontier model leaves behind is a standard… an answer helps you once, a standard upgrades every answer that comes after it

your CLAUDE.md, your skills, your learnings files, your memory setup: this is the layer every future model reads before it touches your work

today, fable writes that layer at a level opus can follow but could never author

run this in every project you care about:

xmlread this entire project and how i work in it

then rewrite my CLAUDE.md as the operating manual a less capable model would need to work here at your level:

the conventions i follow and the ones you’d add the mistakes a weaker model will make in this codebase, named, with the rule that prevents each the quality bar per deliverable, written as checkable criteria, not adjectives what to do when uncertain: the exact escalation rules

then propose the 3 skills that would save me the most hours, and write them in full

the criteria line is the point

a cheaper model can’t invent a quality bar, but it applies a written one fine

next up: pointing that same judgment at the business itself instead of the workspace

2. the consultant audit

fable’s verified edge is judgment on hard, messy problems: on the hardest coding benchmark tier it scores more than double the next model, and the gap widens as difficulty rises

so give it the hardest messy problem you own: your business

open a session with access to your projects, your numbers, whatever context you can feed it, and run:

xmlact as the consultant i can’t afford

audit everything: projects, offers, workflows, pricing, where my time goes

deliver a roadmap i can execute with a less capable model:

ranked moves, highest expected return first per move: why, the exact steps, what done looks like, what a weaker model needs to be told to execute it the three things i should stop doing, with the reasoning written out in full

the deliverable rule carries the value: the reasoning gets written down today, while the model that can produce it is flat-rate

tomorrow opus doesn’t need to be brilliant, it needs to follow a brilliant document

and a roadmap is only as good as the knowledge underneath it, which is move three

3. the second brain run

research is where the extraction runs deepest: long multi-step synthesis is fable’s widest measured lead over every other model

so spend a slice of today on volume: deep research runs on your niche, your competitors, your customers’ problems, the methods you keep meaning to study

then mine every run into an Obsidian vault, the free note app where each note links to related notes, one insight per note

the vault becomes the context every future session reads

full walkthrough of that mining system here:

Machina@EXM7777·Jul 3 ArticleHow to build a second brain with Fable 5I’m going to show you, step by step, how to turn Fable 5 into a machine that knows your business inside out… and ships outputs that look nothing like what everyone else is getting the tool is a…643322.5K1.5M

don’t summarize into one long report… atomize

a hundred linked one-insight notes get retrieved and reused, a 40-page report gets stored and forgotten

the first three moves bank judgment

the fourth spends the thing you actually lose tomorrow: unattended hours

4. fire the goals

fable’s signature capability is holding one job for hours without losing the plot

that endurance is exactly what stops being flat-rate tomorrow, so today it goes to work

two commands in Claude Code, the terminal app the model runs in:

/goal sets a finish line instead of a prompt: you describe what done looks like, and the model keeps working turn after turn while a second, smaller model checks the condition after every turn and only stops the run when it’s met

dynamic workflows are the scale layer: the model writes an orchestration script for your task, and that script runs dozens of subagents in the background, in parallel, cross-checking each other’s findings while your session stays free

the combination is the move: a /goal holds the finish line, the workflow does the fan-out

xml/goal every module in this repo has a test file, the full test suite passes with the complete green run pasted in this chat, and a migration-notes.md documents every change… or stop after 25 turns and paste the failures

two rules make this safe instead of expensive:

demand pasted proof in the finish line: the judge model reads only the conversation, it can’t run your tests or open your files, so the condition asks for the green run pasted, never promised

cap every run: turns or wall-clock, written into the condition, one unattended loop without a cap billed $6,000 by morning

and mind the meter: fable burns weekly limits roughly twice as fast as opus, and only half your weekly limit applies to it in the first place

pick the two or three goals with the most locked-up value, not ten

loops, goals, and 25 ready workflows to steal from:

Machina@EXM7777·Jul 4 ArticleThe Fable Loop Library: 25 Workflows on Autopiloti’m going to teach you how to run Fable 5 on autopilot, using my own library of loops and goals… 25 workflows, each with a prompt and the exact tool it plugs into the method follows karpathy’s…321281K655K

everything so far extracts what fable knows

the last move extracts how it thinks, automatically, for the rest of the day

5. the skill that documents how fable thinks

every time fable cracks a hard problem today, its approach evaporates when the session ends

this move installs a recorder

create the file .claude/skills/extract-approach/SKILL.md:

then wire it into CLAUDE.md so it fires without being asked:

xml## learning law after every non-trivial solved problem, run the extract-approach skill before moving on a solution without its learnings note is unfinished work

now spend the rest of the day working fable hard on your real backlog: the gnarly bug, the architecture decision you’ve been circling

every solve leaves a note behind, and the notes are the distillate: fable’s reasoning, sitting in your repo, readable by every model that comes after

this is the compounding move, and it’s also the one to install FIRST if you only have an hour… it converts all your remaining fable time into permanent assets automatically

[VISUAL: the five moves as a system map, what each one leaves behind after july 7]

after tomorrow

this exact situation will repeat

the pattern of the last year: a frontier model appears, gets re-priced, gets pulled, sometimes comes back, eventually retires

one major model was removed with no notice, resurrected after the backlash, then retired anyway six months later

outputs you don’t save while you have access can’t be reconstructed once it’s gone

every teacher leaves… what you extracted is what you keep

so keep this playbook

the next time a frontier window opens, you won’t spend it on demo apps

quick recap

the test: can a cheaper model redo this tomorrow… if yes, skip it

1. workspace: fable rewrites CLAUDE.md + skills as checkable standards

2. audit: fable the consultant writes the roadmap, opus executes it later

3. second brain: deep research runs, mined into an atomized Obsidian vault

4. goals: /goal + dynamic workflows on your highest-value backlog, pasted proof + hard caps

5. recorder: the extract-approach skill, wired into CLAUDE.md, one learnings note per solved problem

order if time is short: 5 first, then 4 (it runs alone), then 1, 2, 3

the training, the skills library, and the weekly guides that go with this live at weeklyaiops.com