GM

[GM @cleverhoods]

The State of AI Instruction Quality: Q3 2026
https://t.co/vur7Q9xtJy

[ClaudeDevs @ClaudeDevs]

We’re adding support for MCP Tasks and Elicitations in Claude.

Tasks let Claude give plugins long-running jobs such as deep research and report back when they’re done.

[GM @cleverhoods]

@PR0GRAMMERHUM0R real man test on production

[GM @cleverhoods]

@_mctrinh the result

[GM @cleverhoods]

@atmoio well, it's just only a question of time to have models compete against one another in the security landscape.
not yet a ghost in the shell vibes, more of pre-Cyberpunk era

[GM @cleverhoods]

@franagl @ben_sage actually I think @jonlajoiecomedy already did this with "Please use this song"

[GM @cleverhoods]

@ben_sage https://t.co/q488gThClA

[GM @cleverhoods]

@flohoeller man ... thanks for your work, I got a lot of inspiration from it

[GM @cleverhoods]

Over a thousand runs per finding: instructions that name the exact construct got followed roughly 10 times more often than ones that describe the category.

"Handle errors properly in API calls."
"Wrap calls to stripe.Customer.create() in try/except StripeError."

https://t.co/wGdrbO89HV

[GM @cleverhoods]

An architect told me: "The instruction merely influences the probability distribution over outputs. It doesn't override it."

Agreed, and that distribution moves with the wording: naming the exact construct produced roughly a 10× compliance effect in my controlled experiments.
https://t.co/xHIsyEn8G7

[GM @cleverhoods]

@ClaudeDevs any chance to get Fable 5.5?

[GM @cleverhoods]

Finally!!!

[ClaudeDevs @ClaudeDevs]

We're donating Agent Skills to the Agentic AI Foundation (@AgenticAIFdn), the same foundation we donated MCP to.

Since December 2025, 70+ agent platforms have integrated the standard, including ChatGPT, Cursor, GitHub Copilot, and VS Code.

Here's what that means for you:

[GM @cleverhoods]

@ClaudeDevs @AgenticAIFdn oh wow! NICE!!!

[GM @cleverhoods]

@atmoio

[GM @cleverhoods]

@b4dgerr https://t.co/d4yl2TNw1k

[GM @cleverhoods]

@thekitze bro forgot to say "make no mistakes"

[GM @cleverhoods]

@CiprianiRanieri It takes your skills/agents/rules and validates it against 120 research backed rules

[GM @cleverhoods]

@iceccorp it really depends how well you two can work together

[GM @cleverhoods]

@arham_humayun99 my experience goes otherwise. If you want to have DRY, single ownership, SOLID or meaningful testings then you better have some strong steering and progressive disclosure to manage that. Otherwise it will just produce code that is true only for the given task, but fails on overall architecture.

[GM @cleverhoods]

@hakimuddinkika https://t.co/MzfAPOJ1h1

[GM @cleverhoods]

@aditikbuilds https://t.co/MzfAPOJz6z