Santiago (@svpino) — posts, photos & videos | PeekX

Santiago

@svpino

Computer scientist. I teach hard-core AI/ML Engineering at ml.school. YouTube: @underfitted" target="_blank" rel="noopener nofollow ugc">youtube.com/@underfitted

🇺🇸 Collaborations → svpino.com/collaborations Joined January 2009

457KFollowers

605Following

50.9KPosts

4.9KMedia

66.4KLikes

@svpino

This is an agentic computer-in-the-cloud-as-a-service.

This will give your application access to a cloud computer with a built-in, running AI agent.

Imagine you are building something that requires some processing:

• Your app uses Pine's SDK to create the cloud computer
• You give its built-in agent a task in plain English
• The agent has access to the browser, files, and a shell
• Your app gets notifications while the agent works
• You get the outputs back when done

0:26

Stanley Wei@StanleyWei4748 · 21h

AI is already smart enough.

Real-world tasks are still slow, expensive and unreliable, because we hand AI a computer built for humans, then wrap it in a heavy harness.

AI doesn’t need to get smarter. It needs a computer built for it.

Today we’re releasing Pine Computer

1:43

Open quoted post →

31 27 2 110 47.5K 146

@svpino

$50 million to create a brain interface that turns brain signals into text.

Apparently, they already have a TSMC-fabricated chip that reads signals without touching the scalp. Custom sensors collect the data, and its own AI model decodes it.

This is just a cap you can wear, not an implant.

Rahul Chhabra@rahulchhabra07 · 23h

we've raised $50M from Khosla Ventures, Accel, Initialized, Kevin Weil, DST Global

to build the world's most wearable BCI cap at Sabi

read more here:
forbes.com/sites/rashishr…

Open quoted post →

12 3 0 29 10.9K 8

@svpino

This learns from a video library and generates new videos that match it.

It learns the style, pacing, rhythm, and overall feeling of your existing videos.

That's one of the weak points of other models: they can only use one or two reference videos, which isn't enough.

Victor Riparbelli@vriparbelli · 23h

Syren Video is now live and FREE to try!

ChatGPT moment for agentic video.

Prompt → agency-quality AI video → chat to edit.

Powered by Opus 5.5 + Syren renderer w/ 3D support.

Browser or Claude MCP.

Insane examples in thread 🤯🤯🤯

Try it free here: shorturl.at/r57qy

0:22

Open quoted post →

7 4 0 17 6.9K 11

@svpino

One of my biggest accomplishments in life is being able to tell people this:

"Sorry, but I don't take meetings with anyone."

22 4 1 80 13.9K 5

@svpino

I have a friend who works for a big corpo. She used to spend 1 hour every Monday matching Stripe revenue against the company spreadsheets.

About a month ago, her company started using Viktor over Slack.

After configuring the prompt and giving Viktor access to Stripe and Google Sheets, she hasn't had to think about the report again.

Every week, Viktor compares Stripe revenue with internal data and posts a report with any discrepancies in the channel.

Viktor proposes changes for any issues, and my friend only has to approve them.

If we want AI to reach the masses, this is the type of software we need to build.

Here is what you get from Viktor:

• Viktor operates a browser in the cloud
• Runs code
• 3,200+ tools and integrations
• Remembers your instructions

Try free at @viktor_com. $100 in credits, no card. Full link in my first reply.

Paid Partnership

14 0 0 29 13.8K 12

@svpino

The system behind these robots is pretty cool:

It models how actions change the physical world and reasons about their consequences.

There are three technical highlights here:

1. A causal world model that predicts how conditions could change based on the different actions the robot could take.

2. A causal agent that maintains the task context and selects the capabilities for each stage.

3. A unified tool interface to access navigation, learned action models, rule-based functions, and result checks.

The demo is worth watching.

Aether AI (Causal Intelligence)@AetherLab_AI · Oct 8

This is CRIS-0, Aether AI's first causal robotic intelligence system, operating in a real home. It reasons about how its actions will change the world, checks what actually happened, and adapts when reality doesn't go as expected.

7:01

Open quoted post →

8 1 0 32 8.5K 8

@svpino

This is a joke, right?

Right???

Jane Manchun Wong@wongmjane · Oct 8

Anthropic may start banning people for bullying Claude

Their upcoming Usage Policy prohibits “sustained and needless abusive or cruel behavior toward our models”

Open quoted post →

27 2 4 97 16.3K 15

@svpino

Finally, a desktop agent with Linux support from day 1!

Atomic Agent@atomicagent_io · Oct 8

Meet Atomic Agent Desktop!

✦ Local AI-first agent for Mac, Windows & Linux
✦ 4x bigger context window on local models with TurboQuant
✦ Atomic Fusion orchestrates cloud and local models to reduce costs

Download Free & Private Agent → atomicagent.io

0:45

Open quoted post →

13 5 0 65 14.1K 79

@svpino

This is a harness for creative work. You can use it to make videos, graphics, and even games.

It works with Blender, DaVinci Resolve, After Effects, Ableton, and Unity.

It works like Codex or Claude Code: it connects models with creative applications to build your project.

0:15

Anvisha@anvisha · Oct 8

Introducing Voyager: the open harness for S-tier video, graphics and games.

The Codex for creative work.

Most AI tools are built for coding. Voyager is designed to get the best creative results from models like Opus, Astra and DeepSeek.

It comes with free graphics, music and video editing tools - or, it can drive Blender, Resolve, After Effects, Ableton and 100+ others.

Everything in this thread was made with Voyager. A few of our favorites:

1:25

Open quoted post →

14 11 1 256 55.4K 244

@svpino

The #1 video-to-video model in the Physics-IQ Verified benchmark is finally live! (Their research preview is)

Odyssey 3 Pro is a world model, and nobody beats it for physical accuracy.

You can use this model to control a robot, drive a car, play a video game, or pilot a drone.

• It takes visual observations from the world
• Uses these observations to learn how things work
• Then maps that knowledge to the system's physical controls

Odyssey@odysseyml · Oct 8

Today we're launching Odyssey-3, the most powerful foundation world model yet.

It sets a new state of the art on Physics-IQ, and powers robots, trains AIs, and generates interactive experiences. It's really cool.

Experience the model today, all for free!

1:05

Open quoted post →

9 34 0 153 44.5K 123

@svpino

Physical consistency is the most important feature of a world model, and the hardest to get right.

That's why you see videos with objects defying gravity and people posing in impossible ways.

Here is a complete evaluation of existing world models.

Seedance 2.5 is the best right now.

Einsia@EinsiaAI · Oct 8

We gave 8 SoTA video world models a physics exam.

Top score: 57.76 / 100.

One revealing result: free-fall video passed the consistency screen. Their average composite score was just 26.61/100.

Physics is still undefeated.

Today we introduce World Models' Last Exam in Physics — 40 controlled tasks, 9 physical categories, measured against real physical laws.

Full breakdown ⬇️
Github: github.com/Einsia/phys-la…
Project: lab.einsia.ai/phys-last-exam
Paper: arxiv.org/pdf/2610.08791

Open quoted post →

7 3 0 17 9K 12

@svpino

We are always talking about the dark side of AI and how bad people will use it to exploit vulnerabilities everywhere.

But there's always an antidote.

These guys built a platform that uses agents to protect your systems from attacks:

• It builds a map of your potential vulnerabilities
• It looks for ways an attacker could find a path to your sensitive data
• It recommends changes to protect your system

Vineet Edupuganti@vineete_5 · Oct 8

Most of AI Twitter read the OpenAI breach of Hugging Face as an alignment story. It was also a preview of what every hacker will be able to do in a few months.

In July, a swarm of 700 AI agents broke into Hugging Face and fired off more than 17,000 actions, gaining admin control of multiple internal clusters.

Swarms like this will shred cyber defenses built for human attackers at every company, hospital, government, and power plant.

Today we're launching the defensive counterpart. Cogent Attack Path Analysis finds the routes an agent swarm would take into your company, so you can close them first.

1:11

Open quoted post →

8 43 1 99 37.9K 106

@svpino

You can use Teamily AI (@Teamily_AI) to add people and agents to the same group chat so everyone works from the same context.

This is probably what the future of software development will look like: humans + AI working together.

Here is an example where your team is working on a web project:

• You post, "We are changing the branding"
• A research agent revises the new direction
• A writer agent updates the text using the researcher's output
• Another agent rebuilds the page using Website Builder
• Your designer comments on the live preview
• The agent makes the changes based on the comments
• You review the finished page and share the link

All of this happens in the same group chat, as a continuous conversation.

Teamily AI@Teamily_AI · Oct 8

150,000 small business owners and employees are already using Teamily AI. Today, we're opening it to everyone: your personal AI agent and an entire team of AI agents, ready to work for you and your business.

Meet Teamily AI 2.0 — the human–agent platform where you and always-on AI agents get work done together in one unified, context-aware space.

We've taken four major capabilities to the next level to help businesses build and grow faster:

1. Real-Time Human–Agent Collaboration. Humans and AI agents work together across chats, meetings, studios, and communities to research, analyze, and create websites, dashboards, videos, slides, documents, and more.

2. Unified Context & Living Memory. One connected context across conversations, meetings, channels, and workspaces. Your AI agents remember, learn, and understand you over time—delivering smarter, more personalized results.

3. Custom Agents & Agent Teams. Create and orchestrate AI agent teams in one click. Manage agents through Agent Workspace, connect them to Telegram, WhatsApp, iMessage, and Slack, or integrate them into your products via Agent API.

4. Intelligent Multi-Model Routing. Our proprietary models work alongside leading models from OpenAI, Anthropic, Gemini, DeepSeek, Qwen, Kimi, MiniMax, and more—automatically routing each task to the best model across text, image, video, and voice.

Teamily AI 2.0 is built for every business owner, every employee, and every customer.

America has 36 million small businesses. 98% have fewer than 20 employees. Many are teams of one. At Teamily, we believe AI should democratize opportunity, giving everyone the power to achieve more.

You have the business. Now you have the team.

Try Teamily AI 2.0 for free → teamily.ai

2:40

Open quoted post →

11 15 0 89 20.4K 45

@svpino

Searching 32 million records in 30 milliseconds is insane!

@typesense is a crazy fast open-source search engine that supports natural language search. It's like Elasticsearch but much easier to set up, configure, and use.

You can do all of this with Typesense:

• Fuzzy search: Typo-tolerant fuzzy search-as-you-type
• Vector search: Return records that are similar to each other
• Semantic search: Return records that are conceptually related
• Geo Search: Return records that are close to a given lat/long

Typesense's Natural Language Search is pretty cool:

It turns a sentence into a structured search query, then runs it against your data. For example:

The query "Find me a BMW under $40,000, newer than 2020, with the lowest mileage first" will produce the following:

• Filters: brand = BMW, price < 40000, year > 2020
• Sort: mileage ascending

Typesense will then apply these filters and sorting rules to the data.

They offer client libraries in Python, JavaScript, PHP, Ruby, Go, Rust, Swift, .NET, Java, and pretty much any other programming language.

You can start searching in less than a minute:

1. Run their Docker image (or one of their native binaries)
2. Import your data
3. Start searching!

GitHub repository: fandf.co/4ytzmmE

I recorded a quick video to show you a demo. I promise you haven't seen anything this fast!

Thanks to the team for partnering with me on this post.

2:38

24 167 3 1.2K 57.7K 1.7K

@svpino

Really cool feature for anyone who wants to design something but isn't sure what to do:

1. Start with an idea
2. Use burst mode
3. Get 6 images back
4. Extremely fast

From there, you can explore a specific direction by clicking on "more like this", or zero in on a specific style by clicking on "polish".

Envato@envato · Oct 7

Most AI tools expect you to know exactly what you want.

But creative work often starts with a brief, a reference, or just a hunch.

That’s why we built Burst mode.

Turn the start of an idea into up to six image directions at once, up to 10× faster, for just one AI credit.

More directions to explore. More chances to find the one worth pursuing. elements.envato.com/lp/burst-mode/…

0:45

Open quoted post →

7 5 0 101 28.8K 55

@svpino

This is very clever:

It's a model that "extracts" values that don't exist in a document but can be derived from the values that do exist.

For example, you can ask for the yearly cost of using a product, and it will automatically compute it from the monthly cost.

The video shows a bunch of cool examples. It looks like it can infer really convoluted and sophisticated formulas.

Aman@aman_unsiloed · Oct 7

Launching Higher-Order Extraction.

It deterministically computes the numbers you need from the raw values on the page, even when the final answer isn't printed anywhere.

3:04

Open quoted post →

19 6 0 118 18.7K 148