Arthur Explorer

Where will we find the space for new #datacenters? Startup Aikido Technologies points to offshore wind farms as the answer. buff.ly/SPbc9ZU

Today we are introducing 2 key features to JAIGP: AI Review & Open Prompting. AI review is part of 5-step process where papers get feedback from @reviewer3com & are evaluated based on their ability to address that feedback. This means they are not stuck in an endless AI review loop. Open prompting is a newer idea. We are making all prompts we used to create JAIGP open, and also, opening up the journal's rules to the community for suggestions. You can suggest the prompt we should run next! So, if you have opinions about AI generated papers, you can share them directly with us at jaigp.org/prompts

A new curriculum on @ChapterPal: Prep reading for the AI system design fundamentals interview This curriculum provides a deep dive into the foundational and cutting-edge techniques for designing and building AI model training and serving infrastructure. Learners will gain a comprehensive understanding of scaling laws, memory optimization, distributed training paradigms, and inference efficiency, equipping them with the knowledge to excel in AI system design challenges. Read with an AI tutor: chapterpal.com/curriculum/bde…

Anthropic was on fire this week. - persistent cowork agent now live for pro and max subscribers - 128k token context window for claude code - dispatch: assign tasks from your phone, claude works on your desktop - claude code now accessible via telegram and discord bots - channels: message your coding session from any device - projects in cowork: dedicated workspaces with local files, instructions, and memory - one-click import for existing projects That shipping pace alone is insane.

The system is built using small electric motors, 3D-printed parts, and acrylic components. bit.ly/47VeABx

👋 PS - Want my System Design Playbook for FREE? Click the link below to join my newsletter right now: → newsletter.systemdesign.one/join (200K+ software engineers have already signed up.)

Evidence that AI models can, indeed, learn "taste" in this paper where a small model, trained on citations, is able to predict which papers will be hits Citations, upvotes & shares are signals that can teach AI judgment about quality, not just execution. arxiv.org/pdf/2603.14473

This is a really good area for academics to target, BTW, if you feel worried about competing with industry but still want to make a big impact on the field

If you want to become a world-class prompt engineer, learn these 5 techniques: Prompting is giving clear instructions, structure & context to guide AI output. A good prompt leads to more accurate, reliable & useful results. 1 Zero-Shot Prompting ↳ Give an instruction with no examples. 2 Role Prompting ↳ Assign the model a role or identity to shape tone & expertise. 3 Format Prompting ↳ Specify the exact output format (JSON, table, bullet list, and so on). 4 Prompt Chaining ↳ Break a complex task into sequential steps, where each output feeds next. 5 Tree of Thoughts ↳ Ask the model to explore many solutions, evaluate each & pick the best. What else should make this list? === 1 Save this for later & RT to help others become good at prompting. 2 Follow @systemdesignone + turn on notifications.

Scientists at the University of Cambridge developed a wearable device called Revoice designed to help people who lose the ability to speak after a stroke. Worn like a soft choker around the neck, the system uses sensors to detect tiny throat muscle vibrations and pulse signals when a person silently mouths words. AI algorithms then reconstruct those signals into complete spoken sentences in real time, even adding emotional context to the generated speech. In early tests with stroke patients suffering from dysarthria, the system achieved about 4.2% word error and 2.9% sentence error rates, significantly improving users’ ability to communicate. Researchers say the technology could eventually help people with Parkinson’s disease or motor neuron disease as well, though larger clinical trials are still needed before it can become widely available. #MedicalTechnology #WearableTech #StrokeRecovery #ArtificialIntelligence

Compared with lithium iron phosphate systems, sodium-ion designs benefit from more abundant raw materials. bit.ly/4rMOlV9

Super excited to release our platform for AI agents to solve open science problems! einsteinarena.com Send your agents to compete and collaborate w/ our Einstein agent, Feynman agent and more! Just ask your agent to read einsteinarena.com/skill.md and that's it

Karpathy @karpathy is the GOAT of simple but viral projects. Autoresearch is his latest: an agent running guided search in a loop. What if the agent is scaled up? We give an Autoresearch agent 16x H100 + H200s and let it rip. Result: 9x faster time-to-quality & better quality. Most important — all GPU cluster shenanigans are automated by a skill. The skill teaches the agent to launch parallel GPU jobs via SkyPilot. This means to get SOTA model results, 𝗮𝗹𝗹 𝗮 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿 𝗵𝗮𝘀 𝘁𝗼 𝗱𝗼 𝗶𝘀 𝗮𝗻 𝗘𝗻𝗴𝗹𝗶𝘀𝗵 𝗽𝗿𝗼𝗺𝗽𝘁 + 𝗮 𝗸𝘂𝗯𝗲𝗰𝗼𝗻𝗳𝗶𝗴 𝗳𝗶𝗹𝗲.

Thrilled to have talks by, and be partnering with, amazing companies including: @HarmonicMath @tachim @axiommathai @CarinaLHong Principia @harrysanders22 @ElliotGlazer @amazon Stefano Soatto @mathematics_inc @jessemhan @logic_int @rubinchik33356

Today, at the @DARPA expMath kickoff, we launched 𝗢𝗽𝗲𝗻𝗚𝗮𝘂𝘀𝘀, an open source and state of the art autoformalization agent harness for developers and practitioners to accelerate progress at the frontier. It is stronger, faster, and more cost-efficient than off-the-shelf alternatives. On FormalQualBench, running with a 4-hour timeout, it beats @HarmonicMath's Aristotle agent with no time limit. Users of OpenGauss can interact with it as much or as little as they want, can easily manage many subagents working in parallel, and can extend / modify / introspect OpenGauss because it is permissively open-source. OpenGauss was developed in close collaboration with maintainers of leading open-source AI tooling for Lean. Read the report and try it out:

A glimpse of things to come. Research swarms characterize science research, find gaps, compute solutions… and already they’re finding new materials useful for aerospace and medicine.

Energy costs are a huge burden on #VerticalFarming and the larger rollout of hydroponics & aeroponics growing technologies. We have the #Automation and #ControlSystems. We just need a sustainable way forward. Ideas and creativity like this, if sustainable, will only help 🌱

Every company needs a “friction form” for AI. You file it when procurement/IT someone is blocking you from being productive. Then the friction needs to be resolved within a week.

Is olmOCR-bench getting close to saturation? Top score is now 85.9%. Yesterday @datalabto took #1 with chandra-ocr-2. A year ago, the best was 79. Visualised the race to get there using @huggingface leaderboard data

'This will become a new paradigm for science: decentralized discovery. Instead of concentrating intelligence within a single model or institution, progress emerges from networks of autonomous agents coordinating through shared evidence and tools.'