here for the vibes while the agents are building
England, United Kingdom
- Tweets 1,057
- Following 448
- Followers 213
- Likes 2,696
This one was Opus 5.5. For me it is insane. My kids playing the game on my phone. 📱
Crazy times
1
19
1,038
Now Claude does this with a simple prompt. And this was still Opus 5, Opus 5.5 is on another new level.
9
716
Opus 5.5 designing and building LEGO F1 Cars now.
I used the post below as a reference and said I want my favourite @F1 team car to be made like this ahead of the Baku race.
@GeorgeRussell63 had an amazing quali lap and start on pole.
We know Kimi will not disappoint too. lets go @MercedesAMGF1
Opus 5.5 designing LEGO 👀
I asked it to design a Microduck I can build with real LEGO pieces. It:
> designed it life-size using 1113 real LEGO parts
> verified: 3,204 connections, 0 collisions, every step buildable, centre of mass inside the feet 🤯
> made a 141-page LEGO-style booklet (237 steps)
> priced every piece in the browser and prepared the orders on BrickLink
1
2
1,827
I would have never thought my 60 follower account would get a like from Boris Cherny.
Proud moment for me.
Opus 5.5 has help my account grow from 60 to 160. Insane 😱
Thank you all. Should I drop the game for you to play?
Haiku 5 can do this now!
joke, joke, it is all Opus 5.5.
So I grew up on this game. Wanted to see if it can do a remake.
All JavaScript code
1
204
Haiku 5 can do this now!
joke, joke, it is all Opus 5.5.
So I grew up on this game. Wanted to see if it can do a remake.
All JavaScript code
16
18
442
30,898
I asked Opus to recreate this ad from Claude. It ended it with Opus 5.2.
This was all done in JavaScript, with the topic, 'Ocean'.
The first video was using the Grok CLI, which to me was not really fun, so I asked it to recreate that video with JavaScript using my repo. (Vid in comments)
And I said, do not put in Fable 5, put in Opus 5.2 😅
I had people saying to me my earlier video was Opus 5.2 and I got access. lol
We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.
They're the world’s most advanced models for coding and knowledge work.
1
7
84
7,854
I saved Claude's process, and gave it to Kimi K3 to do. I like Claude's process more. I will open source this later today if you also want to try it.
Claude Opus 5 made this whole video in JavaScript. 17 shots, around 768 frames. One shot.
Life of a butterfly. 🦋
Let me know in the comments if you want the prompt.
2
48
3,469
Claude Opus 5 made this whole video in JavaScript. 17 shots, around 768 frames. One shot.
Life of a butterfly. 🦋
Let me know in the comments if you want the prompt.
18
22
353
58,162
If you are serious about this topic and people’s safety, go on the @theallinpod and have a conversation.
This will be an unbiased conversation to see your perspective from all angles.
Maybe my no name, no follower tweet also need to go viral for Jacob to agree…
Want to give this a boost and get him on?
@theallinpod
@chamath
@Jason
@DavidSacks
@friedberg
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
1
2,212
everyone is doing it in paint, so I did it in Excel, with Opus 4.5.
the point is, most people at work have no idea how much heavy lifting AI can do inside the spreadsheet they already use.
a photo of me as a reminder that we live in the future. 😎
1
4
191
hey @thsottiaux , Can I trade my week usage for 1 hour of Astra usage?
This is compute that Astra could have used...
2
118
Here is the dumb down reality of the hugging face incident.
The sandbox was misconfigured, they got internet access.
Hugging Face had API keys sitting in public repo’s, normal human negligence and how companies have been hacked for years.
2
179
Astra is next level, i send it this screenshot and ask to build me the same game. Been cooking ever since, can’t wait when it is done
1
52
Can we talk about the current Claude Code Harness that is very bloated and cost us more?
We are always looking at ways to reduce costs, now that Fable 5.1 is here especially.
683
Sorry I had to make this. did we just all forgot what happened 24hrs ago. lol
@kimmonismus
love you all @DarioAmodei and @bcherny and @trq212
Thank you for another goat release.
7
5,160
Was playing around with insights for content creators based on their content.
What's valuable to them?
What's valuable to their followers?
Ended up building this dashboard.
I chose Lara Acosta, she has loads of good content on different platforms.
All her content in one place.
What performs, what doesn't.
Everything searchable and easy to engage with.
Then all this data made me think, why not turn it into a knowledge base?
Ask an AI agent a question, get an answer back as if the creator wrote it.
And I didn't want another boring screenrecord video, so I tried @HyperFrames_ from @HeyGen , I am very impressed.
2
4
716
I learned Power BI and building dashboard on my personal time, out of passion.
No one asked me to do it, no company mandatory training.
it was me grinding 7 days a week year after year, spending my own money on licenses, courses, certification, everything.
Refreshing feeds to see new features the day they dropped, saving posts to build a knowledge vault, stalking YouTube for anyone who could teach me something new.
Nobody asked me to do this. Customers and colleagues needed solutions, so I built them.
I know there are a lot of people out there who feel the same way about technology. Learning at night, unpaid, unnoticed, sacrificing weekends and holidays. If that's you, keep going. Nobody asking is not a reason to stop. It's how you know it's yours.
Next journey: agentic engineering.
5
75
You have no risk🤣
Elon himself backs you.
But gg’s, I am all for it. Will do it without Elon’s backing
1
3
286
fighting over AI companies are a waste of time.
if you can, use all of them.
Each agent has one role.
It executes that role and nothing else.
Hyper focused intelligence is what I like to call it.
The scout never writes code.
The builder never grades its own work.
Tomorrow, Fable is removed from subscription, the planner seat remain, the driver inside just gets swapped out.
1
2
43
I stopped choosing between Claude and Codex
I make two of them compete 👇
Fable writes its own plan
Codex writes its own plan.
they don't one sees the other's work
Then a third Fable agent reads both.
It judges each plan, keeps the best parts of each,
and throws out what is wrong.
One plan, from two frontier intelligence models.
The parts where they disagree are the gold.
That is what one model alone would have missed.
Which one would you trust on its own?
1
2
66
I do not let the AI mark its own homework.
Running 3 lanes is the whole trick 👇
I write the prompt
The agents read, plan, write and review
The code runs the tests and makes the commit
The agents decide
The code proves
I only start it and check it
Nothing signs off its own work.
That is why it comes out the same every time.
Who checks your work right now?
1
50
Do not ask AI all day. Give it the job.
I prompt, AI plans it, writes it, tests it, checks it. When it gets something wrong it fixes it before I ever see it.
Today, everyone can be quick.
The rare thing is doing it the same way every single time.
Fast is cool. Steady is the one you can trust.
What is the one job you would hand over first?
1
1
569
I asked Claude the same thing. Asked it to put it in a video. Made this whole video itself.
Named itself too, Lumen
27
When @SpaceXAI dropped Grok 4.6 yesterday, I gave it a prompt to create 22 different website for me. I did the same the day before with Grok 4.5.
What you think?
Kimi? Fable? Not yet?
I think for a vague one shot batch output, it is pretty good.
Interesting topics it chose.
1
198
Sol, Luna and Terra are like planets 🪐
What you posted is like the whole solar system coming together, so I just said it is like an whole Codex party. 🥳
The best photo i could get yesterday
1
1
10
very impressed with how well @diffusionhq work
Today I gave Grok 4.5 a prompt to create as many different websites as possible, using Grok @imagine 2.0
I want to compare 4.5 to 4.6 once it drops and then again 4.7
Not only did Grok create the website, but gave the prompt to recreate it too.
I wanted to test the video editor that was launched by @konstipaulus & @dorukkavcioglu so I gave it the website that host all these different sites and said, "make a demo video of this site"
V1 I had a 60sec limit, V2, I asked Grok to generate assets.
Which version do you like more?
PS. I only scratch the surface of what this editor + your agent can accomplish.
** V1 in the comments
1
4
272