The AI discussion thread

[Kaido]

I just binged two of his videos and looking forward to more. Thanks for the recommendation, any other YT specific suggestions?

TONS, what topic are you looking for?

 

[RossMAN]

TONS, what topic are you looking for?

1) anything that's beginner friendly and/or step-by-step friendly.
2) should I register kaido.io and put this entire thread on it?

 

[Kaido]

1) anything that's beginner friendly and/or step-by-step friendly.
2) should I register kaido.io and put this entire thread on it?

Twitter has the best gif summaries tbh!

Here's the basics:

1. All databases are just big spreadsheets.

2. A query is when you ask the database (spreadsheet) for something. In Excel, you can do a CTRL + F to do a find query, type in a formula to do a math query, or draw a bar chart in a data query.

3. The interface is what you use (command line, GUI, etc.)

As far as AI goes:

1. They vacuum up allll of the data on the Internet to create a JUMBO spreadsheet (big ole' database)

2. They have fancy queries called "machine learning" (ML). When you train ML on giant databases, you get a Large Language Model (LLM). When you pre-train it for stuff like human text, you get a Generative Pre-trained Transformer (GPT). So basically, a fancy query! Or rather, modern Microsoft Clippy, haha!

3. The interface can have different modes. Text input is a mode (as well as PDF upload - still text! or programming code text!), as is speaking (speech to text), using an image, using a video, getting data from a database, or getting live data from the Internet or devices. Plus you can now output in each of those modes! So that's called "multi-modal input/output".

Examples:

Take a sample database. Add speech-to-text input, then text-to-speech output. Apply a voice model, set to conversational mode, and use a datacenter for real-time operation. Say hi to Maya on your phone or browser! Ask her how to make chocolate-chip cookies, then try interrupting her to change it to raisins with gluten-free flour:

Sesame

app.sesame.com

What if your input mode was a drawing and your output mode was a video?

https://twitter.com/x/status/1978639812375851014

Or what if your input mode was a drawing and your output mode was a whole computer program?

https://twitter.com/x/status/1982993372748149128

If you're willing to see past the curtain, there are only 3 parts to any AI system, that's it! Data, query, interface! You don't need to know how to code to program in ChatGPT, or how to do web connectors to work in n8n...that's the beauty of it! It's like how artists can use game engines to create beautiful games without needing to master programming...you can start building stuff right now, today, for free!!

https://www.youtube.com/watch?v=NZR64EF3OpA

 

[[DHT]Osiris]

AI in space!

https://twitter.com/x/status/1982794532396900554

View attachment 132731

The power story is VERY interesting!



View attachment 132733

Solar compute power:

View attachment 132732

I do not understand how this is supposed to work. You need cooling for compute, and cooling is very very hard in space. It worked on the ISS because the ISS is an extremely large structure with elements designed for it. I don't see how it could ever be profitable to deploy very large structures for computing resources you can never manually service. Literally putting them in the ocean would be more reasonable.

 

[Kaido]

2-minute anime short made with Sora:

OP answers questions in their comments

15-second clips stitched together in Adobe Premier. Sora image input for character and environment consistency

This project took about 4-8 hours a day generating & editing 20-35s clips. After 4 days, I put them all together in about 2 hours since most of the work was already done. It was all done through the Sora website so absolutely zero budget!


I'll be dropping a discord later this week with workflows, pipelines, and prompt templates. I'll also expose the full prompts for the project as well.

Absolutely game-changing! This will democratize film animation creation for the next generation!!

https://www.reddit.com/r/SoraAi/comments/1oftir8/anime_made_with_sora

 

[Kaido]

Nano Banana (photo) + Wan 2.2 Replace (video) to de-age:

https://twitter.com/x/status/1982952784820334671

19 years ago, they did this with a team of people in X-Men:

 

[Kaido]

Nano Banana (photo) + Wan 2.2 Replace (video) to de-age:



19 years ago, they did this with a team of people in X-Men:

17 years ago, with Brad Pitt in Benjamin Button:

Brad Pitt today:



De-aged 20 years in Nano Banana:

 

[Kaido]

Spot-editing comes to 3D model generation for games, CGI, 3D modeling, etc.

https://twitter.com/x/status/1982855348714381547

Using the ChatGPT Atlas browser to use a prompt to operate the interface. HUGE:

https://twitter.com/x/status/1980732961659187627

 

[Kaido]

AI-generated robot: (3D-generated & then 3D-printed)

https://twitter.com/x/status/1979988656363626743

Inspired by text-to-robot:

Text2Robot: Evolutionary Robot Design from Text Descriptions

International Conference on Robotics and Automation (ICRA 2025)

generalroboticslab.com

 

[Kaido]

A couple fun video generations:

https://twitter.com/x/status/1982776033029108081

https://twitter.com/x/status/1965400951344562237

 

[RnR_au]

A good writeup on Atlas...

ChatGPT's Atlas: The Browser That's Anti-Web - Anil Dash

A blog about making culture. Since 1999.

www.anildash.com

 

[Kaido]

2,000+ n8n workflows:

GitHub - Danitilahun/n8n-workflow-templates: A professionally organized collection of 2,053 n8n workflows with a lightning-fast documentation system that provides instant search, analysis, and browsing capabilities.

A professionally organized collection of 2,053 n8n workflows with a lightning-fast documentation system that provides instant search, analysis, and browsing capabilities. - Danitilahun/n8n-workflow...

github.com

 

[Kaido]

SUPER cool project:

Banano-Shot custom GUI for Nano Banana

Banano-Shot : Easy AI editor, Try 10 Free Edits!

~10/27 Update Feature! Improve UX / UI Improve detail of "Edit mode" Now Available "Remix mode"! Why Banano-Shot? - Because You will get freedom from long Prompt. - 체험판 10장 만들어보기 | Free trial 🤪https://www.banano-shot.xyz/API키 발급 받는곳 | Get API Tokenhttps://replicate.com/account/api-tokens사용방법 URL...

endplan.gumroad.com

Built in Google AI Studio for the base, then Cursor AI & Claude code. Photoshop on steroids!! The camera tool is REALLY cool:



Relighting tool:

 

[RossMAN]

bad news ATOT, brace yourselves ...

Spoiler

kaido.io isn't available, WTF.

domain 1st year renews at
kaido.vision $5.66 $36.56
kaido.club $4.12 $13.90
kaido.global $31.41 $77.75
kaido.art $156.98 $76.94 (WTF?)
kaido.design $9.78 $41.71
kaido.life $2.06 $29.35
kaido.network $4.63 $28.32
kaido.zone $8.24 $31.41
kaido.media $4.63 $36.56
kaido.chat $5.66 $36.56
kaido.solutions $6.18 $25.23
kaido.photography $5.66 $29.35
kaido.guru $2.57 $34.50
kaido.events $11.84 $36.56
kaido.tips $8.24 $25.23
kaido.academy $11.84 $37.59

 

[Kaido]

bad news ATOT, brace yourselves ...

Spoiler

kaido.io isn't available, WTF.

domain 1st year renews at
kaido.vision $5.66 $36.56
kaido.club $4.12 $13.90
kaido.global $31.41 $77.75
kaido.art $156.98 $76.94 (WTF?)
kaido.design $9.78 $41.71
kaido.life $2.06 $29.35
kaido.network $4.63 $28.32
kaido.zone $8.24 $31.41
kaido.media $4.63 $36.56
kaido.chat $5.66 $36.56
kaido.solutions $6.18 $25.23
kaido.photography $5.66 $29.35
kaido.guru $2.57 $34.50
kaido.events $11.84 $36.56
kaido.tips $8.24 $25.23
kaido.academy $11.84 $37.59

STAKE YOUR CLAIM!

 

[Kaido]

Face swap using just 2 reference photos:

https://twitter.com/x/status/1983688522621182297

https://twitter.com/x/status/1984022502025392595

 

[Kaido]

Full character swap:

https://twitter.com/x/status/1984040044106281461

 

[Kaido]

Real-time face & voice change:

This will be as bad of an impact as Sora on society, as far as trust & fake news goes. I already train my IT clients to do verbal passwords for videos to verify identities. The basic process is:

1. Two users go into a secure room with no smartphones or network devices
2. They create a verbal passphrase for both users (ex. "Hey how's the weather where you're at today?" "Just peachy!")
3. They use that on secure business calls

https://twitter.com/x/status/1983846299586683236

 

[[DHT]Osiris]

A few more days down, and I managed to get an arrangement where an LLM is able to reach into our on-prem server running open-webui, and leverage MCPs to do 'stuff' to other systems (so far limited to a couple powershell commands to a test system). In and of itself that's pretty amazing, though. Gotta work on long-term preservation of scripts I've got it building for itself to perform functions, build a library of common scripts (as well as some uncommon ones), and crack open access to the rest of our devices. Automated vuln remediation, here I come!

Finally cracked it:

On-prem open webui instance, using LLM through our own gateway, running an arbitrary command against a remote system. If it can be scripted, our AI can perform it, and it doesn't require me manually setting up claude code on everyone's workstation.