Jay Frey

Nail salon owner → AI builder 🧠 | Went from $600K to $1.6M revenue, now I build software | Practical AI prompting for non-coders | Playbook + templates 👇

[Jay Frey @thetypesetterr]

I really don’t get the hype.. I used Fable 5 and was not impressed.. what am I missing? The “nonstop” workflow? Anything else? I compared F5 with GPT5.5 on xhigh and Opus 4.8 on ultracode. Asked all 3 to find all bugs in the same project. GPT5.5 found 2 more than the others while finding everything those other 2 found… tell me what I’m missing that’s supposedly so “spectacular” about F5.

[Jay Frey @thetypesetterr]

@claudeai @melvynx Still doesn’t fix the lying it does… claiming it did something when it didn’t

[Jay Frey @thetypesetterr]

@petergyang Peter, I have a solution but am in bed. I’ll get it over to you in the morning!

[Jay Frey @thetypesetterr]

@Adidotdev Question. Why IDE when you can use them in terminals directly? Just a curious question.

[Jay Frey @thetypesetterr]

@ziwenxu_ I hope it’s good because they’re Claude dispatch is pretty crappy.

[Jay Frey @thetypesetterr]

@melvynx Have you looked into anthropic:claude-cli for the latest OpenClaw update?

[Jay Frey @thetypesetterr]

@steipete @vincent_koc @openclaw

So, what is anthropic:claude-cli and it's purpose? Will this allow us to use Claude Code CLI meaning the sub, and not the API?

[Jay Frey @thetypesetterr]

@melvynx So, worth the switch?

[Jay Frey @thetypesetterr]

@SabatinoMasala @kenwheeler Yup

[Jay Frey @thetypesetterr]

Yup

[Peter Yang @petergyang]

Both of these can be true:

1. No model is anywhere near as good as Opus for OpenClaw

2. Using Claude Code as a personal assistant replacement is OK but just doesn’t feel as “mine” as my OpenClaw

[Jay Frey @thetypesetterr]

@sama Mmm.. Codex hasn't been getting much love at all, actually

[Jay Frey @thetypesetterr]

@openclaw @vincent_koc @steipete

Just sayin... you know? It's just like... every time...

[Jay Frey @thetypesetterr]

@nateliason @GeeteshParelly Funny how no one has actually simply answered your question...

[Jay Frey @thetypesetterr]

I have this... belief that @steipete and @openclaw team will bring it back. It's only a matter of time.

[Jay Frey @thetypesetterr]

@aa_on_ai @clairevo @openclaw @steipete I've done that before, and I was still spending $250+ per day in Anthropic api fees...

[Jay Frey @thetypesetterr]

@melvynx Yup.. GPT 5.4 is trash, sadly. I've been able to fiddle around and make it better, but night/day difference.

[Jay Frey @thetypesetterr]

@aenigme @kaif9999 So to answer, there is no reason to really use 2.5 > go 2.7. I'm using nearly 12B tokens a month, so for me I'll need to invest in some hardware and run frontier local models as much as possible

[Jay Frey @thetypesetterr]

@kaif9999 So, again...

Why 2.5 vs 2.7? Can you be a kind person and actually explain rather than be arrogant about it? Everyone on here's trying to figure it all out together.

[Jay Frey @thetypesetterr]

@steipete @vincent_koc

This is why @claudeai didn't want to work with @openclaw. Issue.. still API-based meaning usage-based charges.

[Claude @claudeai]

Introducing Claude Managed Agents: everything you need to build and deploy agents at scale.

It pairs an agent harness tuned for performance with production infrastructure, so you can go from prototype to launch in days.

Now in public beta on the Claude Platform.

[Jay Frey @thetypesetterr]

@AnthropicAI @claudeai @steipete Don't you just love how stable #claudecode is now?