@RaedEnergyEng
Account based inUnited States!
About this account
- Account based in
- United States
- Connected via
- Web
X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
مهندس كهربائي مبتكر حلول الطاقة المستدامة
Jeddah
Joined March 2023
- Tweets3.2K
- Following289
- Followers24
- Likes10.2K
Raed Raed retweeted
A lot more is yet to come!
2030 = hard to imagine how much progress we'll have made for humanity by then.
CERN and Mistral AI are building an AI model for physics.
- They're partnering to develop Flare, a foundation model designed for particle physics.
- It will be trained on billions of simulated proton-proton collisions.
- It aims to help reconstruct collisions, identify particles, and flag unexpected signals that could warrant further investigation.
- It could also help evaluate future particle detector designs.
CERN already holds around 1 exabyte of experimental data from the Large Hadron Collider. With the upcoming High-luminosity LHC upgrade expected to increase collision volumes 10× around 2030, analyzing this data will become even more demanding.👀
That's where AI can helps scientists analyze increasingly complex experimental data and accelerate fundamental physics research.
Raed Raed retweeted
DeepSeek V4.1 Flash‑1017 beats GPT‑5.6 Sol and Opus 5 on selected coding and agent benchmarks, at much lower token prices. Super duper impressive!
API pricing per million tokens, uncached input / output:
- DeepSeek Flash‑1017: $0.15 / $0.60 off-peak; $0.30 / $1.20 peak.
- GPT‑6 Luna: $0.10 / $0.50.
- GPT‑5.6 Sol: $4 / $20.
- Claude Opus 5: $5 / $25.
Luna is cheaper per uncached token. Flash is substantially cheaper than Sol and Opus, while beating them on some of DeepSeek’s reported tests.
Key benchmarks:
- DeepSWE: 77.0 vs 74.2 for the previous Flash, 73.0 for GPT‑5.6 Sol and 74.0 for Opus 5.
- Terminal-Bench 4.0: 42.3 vs 31.2 for the previous Flash, 39.9 for GPT‑5.6 Sol and 51.8 for Opus 5.
- Automation-Bench: 60.5 vs 54.8 for the previous Flash, 45.8 for GPT‑5.6 Sol and 50.3 for Opus 5.
DeepSeek also reports faster task completion and shorter outputs. Open weights under MIT, with a 1M-token context window.
So yeah, super impressive price-performance launch!
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
Raed Raed retweeted
Acceleration continues.
DeepSeek-V4.1-Flash-1017 is here
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
Raed Raed retweeted
what DeepSeek just managed with v4.1-flash-1017 is incredible
the 552B MoE model now beats opus 5 on DeepSWE and Automation-Bench while costing just $1.20/M output tokens ($0.60 off-peak) over 20x cheaper than opus 5
i still don't understand how they squeezed this much performance out of the same architecture
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
Raed Raed retweeted
Built it for ourselves. Now every agent I run is in one place.
Live on Product Hunt today, any support is much appreciated 🙏
Claude Code at work. Codex at home. Hermes in the cloud.
Now all in one command center, yours to run.
Harness App is live on Product Hunt today 👉 producthunt.com/products/har…
Free and open source. What's your setup?
Raed Raed retweeted
Vercel CEO @rauchg reveals why the hardest part of building an agent is not writing the code, it’s identity, a wallet, and compute that doesn’t die:
"The hardest part about building an agent, believe it or not, is not actually writing the code for the agent. Getting them the access, the identity, the wallet, these are the harder parts."
"The compute of the traditional cloud infrastructure was not really designed for this agent era. It was short-lived compute."
"Vercel Daemons, this primitive for backend compute so that something can run for as long as it wants and respond to events. It’s like an agent that lives forever."
"Drives. As these agents pick up new work, they need a workspace. Think of it as like the Dropbox for agents."
@vercel
Raed Raed retweeted
Samo Burja reveals how Nvidia went from making John Carmack happy to building the GPU hardware that made the 2022 AI boom possible:
"It didn’t even start as a video game graphics company. The core idea was that there would be a new future in compute, and it was parallel processing. Since it was the 1990s, making John Carmack very happy was a way to monetize building a future of compute."
"When we talk about neo clouds, almost all of those are formerly crypto companies that had bought Nvidia GPUs."
"Jensen was right there providing the first custom designed GPU rack that was for AI. This is all pre-2022. 2022 would not have happened were it not for the hardware."
"Maybe he made multiple bets, but it was overdetermined one of those bets would work, and that’s the magic of a live player."
@SamoBurja @bismarckanlys
Raed Raed retweeted
Tesla has pulled the Optimus Gen 3 renders from its Android app. Version 4.61.6-4647, out this week, no longer includes any of the nine robot images. That covers the Gen 3, gold and original Optimus. The Gen 3 files first showed up in 4.60.0 on Aug 27.
Tesla’s Android app has contained Optimus renders labeled “gen3” since August 27.
We compared four app builds: higher-resolution versions arrived September 5, and the latest release contains the same files.
Compared with the gold variant, the “gen3” render covers the exposed mechanisms around the hips and lower legs.
A glimpse of the next Optimus? Possibly. These are app assets—not confirmation of the final hardware or a working robot interface.
Raed Raed retweeted
🛸Vanadium dioxide (VO2), owing to its reversible insulator-to-metal phase transition, is a promising material for tunable terahertz (THz) devices. In this work, a compact VO2-based terahertz absorber is numerically proposed to achieve thermally tunable ultra-broadband absorption. The proposed single-layer metasurface employs a fourfold-symmetric integration of cross-shaped, circular-ring, and square-frame elements to induce multiple coupled resonances and their spectral overlap, providing a design strategy for broadening the absorption bandwidth while maintaining polarization-insensitive operation.
Raed Raed retweeted
Cezar Chirila's Picothermal is an open source, USB-C thermal camera that turns your smartphone into a 160×120 thermal imager without installing an app. hackster.io/news/picothermal…
Raed Raed retweeted
NVIDIA has released Nemotron 3.5 Super VL, an open model that combines reasoning and tool use with text, image, document, and video understanding.
It has 120B total parameters, activates 12B per token, and supports up to a 1M-token context window.
Weights are available on Hugging Face, with hosted access through OpenRouter and deployment through NVIDIA NIM.
For developers building agents, the practical addition is handling visual information and taking the next step within the same model.
Raed Raed retweeted
”The result is a 40x error reduction on our Willow processor vs. previous methods, achieving a state-of-the-art 99.99% logical fidelity.”
This is a massive improvement for usable quantum computing.
huge congratulations to the folks at google
Some magical progress towards "magic state" in quantum!
Our @GoogleQuantumAI team has demonstrated "magic state cultivation" on physical hardware for the first time. The result is a 40x error reduction on our Willow processor vs. previous methods, achieving a state-of-the-art 99.99% logical fidelity.
It's another step towards useful, error-corrected quantum computers to help solve the hardest problems. The result is published in Science today, watch the video below to see how it works
Raed Raed retweeted
NEW ROBOT FIGHT: A model specialized in robot boxing fights!
Called RoboStriker, it is a method for autonomous 1-versus-1 boxing between two 29-DoF Unitree G1 humanoids, from a Chinese consortium (Shanghai Jiao Tong University, Shanghai AI Laboratory, Shanghai Innovation Institute, HUST, USTC).
It is a three-stage framework:
1. a motion tracker is first trained by imitation on about 30 minutes of professional-boxer motion capture
2. those skills are distilled by teacher-student into a 32-dimensional latent action manifold
3. two fighters are trained by self-play.
Core of the method: direct self-play in raw 29-DoF joint space causes catastrophic balance failures.
RoboStriker first distills about 30 minutes of pro-boxer mocap into a 32-dimensional latent skill manifold and plays the game there.
-> To me, this is also proof that URKL is in no way autonomous, no matter what they claim.
Raed Raed retweeted
Asana CPO @ArBose argues that today's AI chat interface is fundamentally broken, predicting a future where AI agents know what you need before you even ask:
"I don’t think that many enterprises actually have long horizon agents running today. Almost everybody is using the chat tools to create content or to summarize meetings. In a year’s time, it’ll be the norm for long horizon complex workflows to be agentic."
"I think the current user experience for engaging with AI is broken. It’s a chat window. You have to come up with the creativity to ask it questions or provide it with a blurb of text and context and hope that the result is good."
"Proactivity is gonna be the big sea change, and it’ll take the impact of AI from a small percentage of people who actually use it today on a day-to-day basis to the mainstream."
@asana
Raed Raed retweeted
@Cisco presented 400G bidirectional optics that can boost speeds over existing 100G fiber, potentially avoiding costly recabling. #NFD41 #LinkedIn linkedin.com/pulse/400g-bidi…
Raed Raed retweeted
TSMC earnings note live in CS Atlas. Crossed with our WFE model, TSMC earnings forecast, their CPU commentary and their power shell checks vs. ours.
atlas.creativestrategies.com…
Raed Raed retweeted
Introducing Span-02: anytime reasoning for decisions.
Reasoning models make you pick an effort level and wait. Span-02 takes a deadline and always returns its best decision so far.
It gets better the longer it thinks. With reasoning off, it beats Jev at 1/4 of the price, and runs at 1/3 the latency of decision models at its level.
It can:
- Answer with numbers and scores, not just probabilities and choices
- See images and screens (multimodal)
- Take a configurable reasoning time
And it’s fully open weight!
Raed Raed retweeted
one of the more interesting things happening in AI right now is the shift from personal assistants to agents that can actually work together
@zenzap lets businesses create AI agents, connect them to existing tools and automate workflows
🤝 Paid partnership
AI is everywhere, but walk into almost any company and work hasn’t changed
Meet @Zenzap_chat, the Multiplayer AI platform.
Now, people and proactive AI agents are in the same group chats,
automating workflows with shared context across business tools and teams
Everyone can bring their team and create AI agents in 90 secs (even from your phone). CISOs and CIOs love it
Thousand of teams from our early access program already transformed their day to day
AI agents have joined the team
Raed Raed retweeted
Robots are taking over construction sites!
Wall finishing is one of the most labor-intensive jobs in construction. Applying putty, sanding surfaces, and painting still require hours of repetitive manual work.
China's HX Precision has developed the HX33-W, a 3-in-1 wall finishing robot that handles all three tasks.
The robot automatically applies and smooths putty, sands walls, and paints surfaces, navigating construction sites without continuous manual control.
According to the manufacturer, it delivers 4–6× higher productivity than traditional manual work, reaches heights of up to 3.3 meters, and operates for 6–8 hours on battery power.
What's interesting is how specialized these machines are becoming. Instead of trying to automate an entire construction site, they focus on repetitive tasks where labor costs, consistency, and productivity matter.
Construction has been notoriously difficult to automate. Robots like this could start changing that.
~~
🤖 Interested in robotics and AI? → 22astronauts.com