SciWorkflow

SciWorkflow-Bench: A Long-Horizon Agentic Benchmark for End-to-End Scientific Research Pipelines in Biology x Materials Science x Coding and Engineering

[Chakra @chakra_ui]

Floating Panel just got shipped ⚡️

➡️Get started: https://t.co/AVohvFlDzM

[CyberRobo @CyberRobooo]

MindOn's new demo, One model(Mind-0)drives humanoids + dual-arm robots in full logistics workflow--all trained on human-centric data only.

Shared intelligence makes it possible for complex logistics operations to run through intelligent automation.

[Qingxu Zhu @QingxuZhu]

We’re sharing a new MindOn demo.
Two humanoid robots and two dual-arm robots complete a shared logistics workflow—from picking to packing.
One model. Multiple embodiments. No robot-collected data, only human-centric learning.
It generalizes across bodies and coordinates real-world manipulation.

[SciWorkflow @SciWorkFlow]

Part 2 Update — One Man Marathon (SciWorkflow-Bench)
Quick honest status tonight:
We spent the weekend building out the 15 robotics benchmarks in a separate UI. 10 of them are fully working and connected. The other 5 still need more polishing.
While doing that, we accidentally broke/destroyed the original main dashboard (the first big benchmark page with the 10 core tasks). So tonight we focused on unifying everything again — trying to bring back the original 10 benchmarks + keep the robotics section intact.
Current state:
• Main dashboard is partially restored with 10 benchmarks + working Grok grading modal
• Robotics page (15 benchmarks) is still there but needs cleanup
• Grading is working on the main benchmarks
• Still need to fully merge them cleanly without breaking anything again
It’s been a grind. Lost some progress on the original page but made real headway on the robotics side. Balancing family, job applications, and this build at the same time.
Tomorrow’s goal: Get the unified version stable, deploy to Render, and start the pitch deck.
One man marathon continues. Slow but steady.

[SciWorkflow @SciWorkFlow]

@0x_fokki @PaulPolino

[Fokki @0x_fokki]

https://t.co/3KfBj1EuKl

[KuphDev @KuphDev]

@SciWorkFlow haha "research" or "fun livestream toy" 🤷‍♂️

Either way I'm having a great time!

[SciWorkflow @SciWorkFlow]

@KuphDev I LOVE this project!!!!!!

[SciWorkflow @SciWorkFlow]

@KuphDev Put the green duck on target

[SciWorkflow @SciWorkFlow]

@KuphDev I love this kinda of research

[KuphDev @KuphDev]

I built a robot that random people on the internet can control by sending it tweets.

This project was both easier & harder than I thought. Using AI felt like having superpowers.

But hardware is still hard.

• Broken servo that took weeks to replace
• Weak 3rd joint that needed a counter-spring
• 9-year-old Craigslist Frankenstein desktop
• 1,000 ACT policy training episodes

After many late nights training models, and a skyrocketing power bill, it actually works!!!

🔴 It's live 24/7 right now!

This was just the MVP though.
The next robot is gonna be even more exciting, and maybe a little dangerous 😈

Reply to the pinned post and send it a command if you wanna test it yourself 🐤

[SciWorkflow @SciWorkFlow]

Project Pitch SUBMITTED!!!!

[SciWorkflow @SciWorkFlow]

@AltcoinDaily Yo!!!

[Horizon @horizon_trade_x]

https://t.co/Bibca2ADfE

[Andrew Bolis @AndrewBolis]

Most people still prompt like it’s 2022.

Here’s how to go from basic to expert-level:

[ bookmark 🔖 this post for later ]

Level 1: Surface Prompts

- Zero-shot prompt: Just ask without examples and hope for the best.
- One-shot prompt: Provide one example to get slightly better results.
- Few-shot prompt: Share multiple examples to guide the answer.
- Easy tasks: Summarize, rewrite, brainstorm, explain like I'm 5.

This is where most stop. It's quick, but basic.
You get generic answers, not high-quality output.

Level 2: Real Work Zone

- Role: Tell the AI who to be and how to sound.
- Tone and style: Define the voice, clarity, or formality.
- Plan → Act → Summarize: Direct the process.
- Define the task: Be specific about what you want.
- Add constraints: Set clear limits and boundaries.
- Provide context: Share background, audience & restrictions.
- Temporary chats: Use ChatGPT without its memory of you.
- Define output format: Bullets, tables, or any structure.
- Tool policy: Turn web browsing on or off.
- Share examples of quality outputs: Set the standard.
- Memory management: Keep projects organized.

This is where quality improves.
You get targeted, practical, and useful results.

Level 3: Where the Magic Happens

- Pick the right model: Select the best tool for the job.
- Thinking vs Fast: Decide if you want thorough or quick answers.
- Reasoning instructions: Tell the AI to think step-by-step.
- Chain-of-Thought: Guide logic instead of just giving commands.
- Iteration loop: Review, revise, and improve responses.
- Problem-solving: Focus on the 20% that gets 80% of results.
- Combine role, context, examples & revision for expert-level output.

The deeper you go, the better your results get.

📌 Get Advanced ChatGPT Guide (free): https://t.co/kOBWfKrBaX

👉 Follow me @AndrewBolis for more and 🔄 Repost this to help others use AI

[SciWorkflow @SciWorkFlow]

@Em_Nomadic Ok

[Emerson S @Em_Nomadic]

This is still one of the cleanest entry points into physical AI for builders.

NVIDIA’s JetBot open-source, built on Jetson Nano, programmable entirely from your web browser.

Under $150 in parts as an add-on to Jetson Nano. Runs AI collision avoidance, object following, and remote control out of the box. All tutorials included. All code open source on GitHub.

The barrier to building and learning on real AI hardware has never been lower.

If you’re curious about physical AI and want hands-on experience rather than just reading about it this is one of the most accessible starting points that exists.

https://t.co/7ZDB2eFaWC

[MTS @MTSlive]

SITUATION DETECTED: Midjourney has announced their first hardware device, a full-body ultrasonic medical scanner.

[Ritvik Singh @ritvik_singh9]

Introducing ABC: open data, training, and infrastructure for robotics.

We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.

@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall

[Idea Browser @ideabrowser]

This sounds dumb…

You should celebrate your failures.

You built, shipped, tested, learned.
1000x better than doing nothing.

The next failure will be faster.
Or it might even be your success.

The only way to know is to keep going.