fable security restrictions for bad ideas

worth it

Inspired by this, I had codex build a tool that I implemented into my Claude settings for the warning and an easy automatic handoff process.

We heard you wanted to use Codex rate limit resets on your own time. Starting today, we’re rolling out the ability to save rate limit resets to use later. We’re starting Go, Plus, Pro, and Business users with one free reset:

the most important intersection of blockchain and ai is interoperable memory. it’s not finance, payments or money. the models and harnesses are constantly changing. i want to bring my history and preferences when i switch harnesses every 2 weeks.

claude code but the model is aware of usage limits and adapts when you get close to hitting them.

my only job now is to make sure my laptop battery doesn't die

stuff that never worked suddenly work

you can't prompt fable 5 the way you prompt the other models. it's a different beast. i've had almost a full day with it, here's my main takeaways: 1. pick your hardest task. anthropic says easy tasks are a waste of tokens. 2. control depth with the effort setting, not bigger prompts. anthropic says fable 5 on low/med beats opus models on xhigh. 3. delete old rules and skills — instruction-following is now so strong that over-specifying actually makes it worse. 4. give it the why not the what — it does better work when it knows the goal, not just the task. 5. it loves to check its own work and verify. tell it what success and evidence looks like. 6. it also loves using subagents. lean into it, it saves time and money. and it does a good job of creating parallel tasks so none of the subagents block eachother. 7. ask it to document its learnings and will self-improve. if you turn off claude code’s memory system, prompt it to write it’s learnings after each turn.

my brain is so tired of reading. coding harnesses should show, not tell.

i have been resisting the urge to tweet about how amazed i am with fable 5 because it’s just so obvious and we don’t need yet another tweet about it.

Am I the only one who thinks git worktrees are fucking confusing?

trade offer: - design services - promotional design concepts - build you a polished working app - my soul - literally anything else for - knicks game 4 tickets

my biggest issue with fable is the need to continue reminding myself to be more ambitious with how i am using it.

ok ok listen: ask fable 5 to create a workflow to run a design audit. and ask it to create visuals for before and after the changes it proposes because you're tired of reading. trust me, bro.

"do a design audit" is my "should i walk or drive to the carwash"

can we please add a row for design to these

how to use loops for frontend and UI: loops won't automate design, but it will help get the request right on the first implementation attempt. agents still suck at interpreting design requests and, worse, one-shotting spacing/sizing, behaviors, animations, transitions and states. the core of every good loop is clearly defining success. so for every design loop, you need a visual reference. if you're designing something new, you can start with a screen recording of an app you like. record as many interactions as possible. every click, hover state, error state, etc. or if you're working within an existing design system, you can explore figma mockups with agent. the most important thing is to prompt with a pixel perfect reference: a recording, screenshot or figma mockup(s). you can share another app and ask it to adapt the visual design or interactive behavior to your existing app. you should explicitly ask it to translate the reference and adhere to your design system (formal or informal). the second most important thing is in your prompt, before the loop begins, you ask the agent to interview you to fill in the gaps to fully understand your intent. a tool like claude code or codex will ask questions like "what should happen when a user does X" to cover edge cases. a few back and forths will build the context to help cover the important details you forgot or didn't think to mention. lastly, before starting the loop, it's absolutely critical you describe success. "verify with computer use everything looks and behaves as intended" or something along those lines. you are essentially offloading the tedious review, and it will check its own work on a loop until it matches your design brief.