Lost Son

[John Galt @dadiani_george]

Spot on.

[Alpha Prime @AlphaaPrimee]

5 Workouts for Huge Chest:

-Thread-

[TheArtOfMan @TheArt0fMan]

Asthetic Workout. Save and try these.

[Shann³ @shannholmberg]

how an ai writing harness works, visualized

1. create the skill that loads your writing .md files, voice-dna, forbidden patterns and examples

2. when you call the skill or prompt your agents to write, it loads in the skill and the reference files

3. it drafts the content, then spins up eval loops from other subagents to check the draft against your guidelines

4. your turn to evaluate, edit what you want, always hand edit before you ship (avoid slop), then either publish yourself or hand over to your orchestrator agents

5. save the approved content to a database, this becomes your reference library for future drafts

always stay under 50% context window to avoid hallucinations and quality drops

[Ruben Hassid @rubenhassid]

I absolutely hate AI writing. And my entire job is to write about AI. So I keep a file of the new em-dash to recognize AI writing instantly:

Everyone learned the em dash trick.
AI is trained to not write with em dashes anymore.
Now you can get accused of using AI if you write in perfectly structured sentences.

These are some of the dead giveaways of using AI:

1. "That's not X. That's Y."
"That's not compliance. That's stalling."
Say the second half only. "They're stalling."

2. Two short bits stuck together.
"Fast. Simple." "No fluff. Just answers."
Pick only one to write like a real person.

3. "Less a hammer, more a scalpel."
Visualizing two pictures with no advice. Directly say what to do instead.

4. Clapping for itself.
"And that matters." "That's the part everyone misses." "Which is exactly the point."
Delete them all. You lose nothing. I promise.

5. "It's the Excel of AI agents."
Only works if I know both things. Usually I don't.

6. Warming up before talking.
"Here's the thing." "Let me be clear." "The truth is."
Start one sentence later.

7. Always three things.
"Faster, cheaper, smarter."
Real life gives you two reasons. Or five. Never always three.

8. "5 to 10 minutes."
A range means you never did it. Just say 7 minutes.

9. The ending that repeats the whole post.
"In short…" "At the end of the day…"
Just stop typing. End on a high note.

Now that you know how AI writes. Here's how to actually stop AI from doing all nine:

Step 1. Type only these two words in your prompt.
Open chat & write: "Use ASD-STE100."

That's the writing standard for plane manuals. Short words, short sentences, because a confusing sentence there can kill someone. Claude already knows it. Those words remove 80% of the bad writing. Works on ChatGPT, Gemini & Grok too.

Step 2. Get the skill so you never type it again.
I built the full version with every rule inside. It's free: https://t.co/rFXoSKBlNd

Download the zip. In Claude, click Customize on the left. Click Skills. Click Upload a skill. Drop the zip in. Type /ste to use it.

Step 3. Turn it off for fun stuff.
I asked it for a poem. Unreadable. I asked it for a LinkedIn post. It sounded German.
Use /ste for guides, steps, explanations, emails, documents. Turn it off for anything with a heart.

Step 4. Add 4 lines to your claude .md.
"follow Zinsser's four principles of quality writing: 1. Simplicity 2. Brevity 3. Clarity 4. Humanity"

/ste gives you the first three but forgets about the fourth. This prompt puts it back on the track.

Step 5. Make a forbidden .md.
Copy the 9 patterns above into a file. Tell Claude to check every draft against it.

Then add a new line every time you spot one. Your file gets better every week. Mine certainly does.

Step 6. Start a new chat more often.
If your chat is long, the writing gets worse. It repeats itself & goes back to the pretty sentences. When it feels half full, open a fresh chat & paste your files back.

The skill is free, right in your inbox: https://t.co/rFXoSKBlNd

[Anatoli Kopadze @AnatoliKopadze]

Anthropic engineer:

"You don't need better prompts, you need graph engineering that makes agents remember everything."

In 30 minutes two Anthropic engineers show how agent-building has evolved.

Then they build agents from scratch that recover from failures, remember across sessions, and improve themselves every night through a process they call "dreaming".

Watch it, then read the guide below on how to build a system that improves itself.

[Lost Son @rawle_charles]

@hdfilmkanalii Looks amazing, need to find subtitles to truly enjoy it

[Lost Son @rawle_charles]

@Raindropsmedia1 She is exhausting! You would never have peace with her

[Lost Son @rawle_charles]

@Li_Rois_Artus @gotrice2024 Deserved is a strange word. Who is to say what someone deserved?

[victóry | Educator @egbokavictory_]

How the rich train their kids.

[Suzie rizzio @Suzierizzo1]

Who remembers that the Army swore in four tech Executives as Lieutenant Colonels and these men are now one rank away from the top! Military members normally have to rise through 16 ranks to get to this level. This is definitely not earned Valor it’s Stolen Valor & a Disgrace!

[kevin blue @kevinblue345]

white people like to lie to themselves that they abolished slavery and that they freed the black slaves. Black people freed themselves.

[Meidas_Charise Lee @charise_lee]

We knew he didn’t do it‼️👇🏼 Trey Reed RIP

[ᗰᗩƳᖇᗩ @LePapillonBlu2]

Trump doesn’t want you to see this.

[Where We Going, We Don't Need Britches 💙💛 @RPHutch1975]

Good morning, Bluesky
Blue Dad, [A], Apocalypse
Art by Fernando Pasarin

[MAGA Cult Slayer @MAGACult2]

Stop normalizing this. Stop covering both sides of it. Start calling out the technocratic oligarchs now. They are the ones implementing Project 2025 and that’s exactly what this is, a dictatorship.

[Meidas_Charise Lee @charise_lee]

America is a joke to the rest of the world‼️

[Lost Son @rawle_charles]

@mermaidmamamags @GavinNewsom