Dany

learning AI in public. testing what works, ignoring what doesn’t. occasionally philosophical about it.

[Dany @danysvnt]

https://t.co/cZpIUKF54F

[Dany @danysvnt]

https://t.co/cZpIUKF54F

[Dany @danysvnt]

HE'S MAKING MONEY WITH THE FIVE APPS YOU'VE BEEN RESEARCHING ALL YEAR

fifty seconds. five tools. one guy running an entire business on them.

grok for what's actually trending on X this hour. voice dictation that turns a mumbled sentence into a finished email. chatgpt image 2.0. notebooklm. claude sitting at number one.

you already have access to all five. two of them are probably open in another tab right now.

there's a second in the video where grok pulls up a post saying the people winning right now aren't the smartest, they're the fastest to ship. he glances at it and keeps working.

honestly that's the entire lesson. same tools, same tabs, different relationship with starting.

pick one of the five and point it at something real before you close this app.

[Dany @danysvnt]

https://t.co/cZpIUKF54F

[Dany @danysvnt]

I ALREADY KNOW YOU WON'T WATCH THIS ONE. 22 HOURS, RIGHT?

freecodecamp put up a 22 hour course called GenAI Essentials two days ago. free. 43,000 people got there before you did.

and you're already doing the math on whether you have the time for it. you do. you've spent it already, just in pieces.

a video about embeddings on a tuesday. half a thread about agents on the train home. a reel on prompting you saved and never opened again. every piece true, none of them touching.

honestly that's the whole problem. you never skipped the learning. you did it in fragments too small to turn into a skill.

the guy who's ahead of you has the same tab open that you do. he sat down once.

stop making excuses. start it tonight.

[Dany @danysvnt]

@0xForge daaaamn

[Dany @danysvnt]

@HiddenLore0 crazy how life-changing the right connections can be..

[Dany @danysvnt]

I ALREADY KNOW YOU WON'T WATCH THIS ONE. 22 HOURS, RIGHT?

freecodecamp put up a 22 hour course called GenAI Essentials two days ago. free. 43,000 people got there before you did.

and you're already doing the math on whether you have the time for it. you do. you've spent it already, just in pieces.

a video about embeddings on a tuesday. half a thread about agents on the train home. a reel on prompting you saved and never opened again. every piece true, none of them touching.

honestly that's the whole problem. you never skipped the learning. you did it in fragments too small to turn into a skill.

the guy who's ahead of you has the same tab open that you do. he sat down once.

stop making excuses. start it tonight.

[Dany @danysvnt]

https://t.co/cZpIUKF54F

[Dany @danysvnt]

@0xForge https://t.co/HEsip7h1Kq

you are welcome!

[Dany @danysvnt]

YOU ALREADY THINK YOU KNOW THIS ONE. THAT'S WHY YOU'LL SCROLL PAST IT

it's a 60 second video and it changes how you talk to every AI you touch, not just one of them.

you know the loop. you type what you want, get something almost right, type "make it better", get something worse, and twenty minutes later you're rewriting the same sentence for the fifth time.

that loop isn't the model being dumb. it's the prompt.

ok so the fix takes about a minute:

1. open a new project in chatgpt
2. drop one master prompt in as the system instructions
3. when you need anything, go there and describe the task in plain words
4. it hands you a finished prompt, you paste that into whatever tool you're actually using

works the same whether you're in chatgpt, claude or midjourney. the project doesn't care where the output goes.

the reason it works is stupidly simple. you were going human to AI. now it's human to AI to AI. that middle step is the part nobody adds.

the gap between people getting great output and people fighting with the same chat isn't access. you both have the same tab open. one of you spent a minute on setup once.

the actual prompt file is in the first reply. take it.

[Dany @danysvnt]

https://t.co/cZpIUKF54F

[Dany @danysvnt]

YOU'LL SKIP THIS BECAUSE THE VIDEO IS 16MIN LONG, NOT BECAUSE IT'S USELESS

I promise you won't regret it. by the end you'll know at least three times more than you know right now.

sixteen minutes. that's the entire price. and your brain is already hunting for a reason not to pay it.

what's actually in there:

1. cowork running multi step projects across your own files, not one prompt one answer
2. claude code building real working tools without you writing a line
3. artifacts you can publish as a link and hand to a client the same hour
4. claude design turning messy notes into a landing page in your brand colors
5. excel cleaned up, formulas fixed, summary tab built for you
6. claude living inside excel, powerpoint, word and outlook so you never leave the app
7. decks from a rough outline, one idea per slide, export ready
8. skills, so every deck comes out on brand without you asking twice
9. projects plus connectors pulling drive, gmail and slack into one place
10. a prompt that finds contradictions buried in a hundred page report
11. a chrome agent clicking and filling forms while you watch
12. dispatch, running work on your desktop while you steer it from your phone on the train

now the part everyone pretends not to care about. this is what the work sells for right now. not a promise, just market rates:

one automation build: $1,500 to $7,500 an agency delivering the same thing: $5,000 to $10,000 keeping a few of them alive: $2,000 to $8,000 a month freelance rate for this work: $90 to $150 an hour a full time AI automation engineer in the US: $90k to $140k a year

nobody is paying that for genius. they're paying it for the twelve things in this video, pointed at someone else's problem instead of your own.

that's the uncomfortable bit. none of this is exotic. it's the floor. it's the base layer that everyone in this field will simply have in twelve months, and right now most people are actively hiding from it behind "I'm doing fine without it."

fine is the most expensive word in tech right now.

you'll tell yourself you'll watch it later. later is where good videos go to die. the person who watches tonight isn't smarter than you, he just didn't spend twenty minutes negotiating with himself over pressing play.

this is the highest return investment available to you and it costs attention, not money. nobody hands you the skill. you sit down, you watch, you build one thing with it.

full focus. no second tab. sixteen minutes.

save this and actually watch it tonight.

[Dany @danysvnt]

https://t.co/cZpIUKF54F

[Dany @danysvnt]

@0xFiseo this article is so underrated

[Dany @danysvnt]

@mccaIln damn, and thats just the beginning

[Dany @danysvnt]

@kyryusha52 SHE'S BEAFTIFUL....... god forgive me this 🙏🙏

[Dany @danysvnt]

@0xBackwood thats actually so sad :(

hope this is gonna be helpful for someone

[Dany @danysvnt]

PLEASEEEE STOP SLEEPING ON CLAUDE SKILLS

save this. you'll want it the next time you watch your limit die at 2pm.

I spent about three weeks pasting the same paragraph into every new chat. how I write, what I hate, what the output should look like. every single time. and then wondering why I ran out of messages by the afternoon.

turns out the whole thing is a folder with one markdown file in it.

here's the part that actually matters and nobody says out loud: Claude only reads the name and description of a skill at startup, something like a hundred tokens. the full file loads only when the task matches. so twenty skills sitting on your machine cost you almost nothing until one of them is actually needed.

your instructions stop living in the chat and start living on disk. that's it. that's the whole feature.

the four that changed my week:

1. skill-creator. you describe a workflow out loud, it interviews you and writes the skill file for you. build this one first, it makes every other one easier.

2. caveman. makes Claude drop the filler and answer in fragments, keeps code and errors exactly intact. the author is refreshingly honest that it mostly saves output tokens and much less on long coding runs. still free money if you're on limits.

3. grill-me. interviews you about your plan before it writes a single line. sounds annoying for about four minutes, then you realise you've stopped rebuilding the same thing three times.

4. your own voice, written down once. nobody builds this one and it's the highest return of all of them. one file describing how you actually sound. everything after that inherits it.

there's an official repo from Anthropic and a searchable directory, both free, I'll drop them in the replies.

you don't need forty skills. install two, write one yourself, and you're already using a different product than everyone complaining about limits.

[Nate Herk @nateherk]

https://t.co/YVrssGik6q

[Dany @danysvnt]

@0xBackwood it seems like we need a looot of time before that actually happens... just my opinion

[Dany @danysvnt]

@0xYaYaYa ahahahhaahha so true

[Dany @danysvnt]

@0xAtlas072 is it actually true that kimi k3 is that powerful?

[Dany @danysvnt]

THIS IS THE EASIEST WAY TO MAKE MONEY IN 2026 AND IT COSTS ZERO

90% of you will skip it. not because it's bad. because it's free, and free never feels urgent. put a $997 checkout page on it and half this timeline would be watching it tonight.

he's pointing at a 5 hour n8n course sitting on youtube for free. voice agents, content automations, full client ready builds. agencies charge 2k a month to maintain systems you'd learn to build there in a weekend.

that's the whole gap. not talent, not luck, not being early. it's people who watch and build vs people who wait for a shortcut in the replies.

making money is not hard, it's boring. learn one thing properly, sell it to someone who needs it, repeat.

save this and come back when you're done looking for the easy version.

[Anatoli Kopadze @AnatoliKopadze]

https://t.co/eq0oL4fIaR

[Dany @danysvnt]

@0xForge AHAHAHAH thats crazy

i`d hire them all tbh