Offline sandbox implementations in eval frameworks are all so different. Cause the agent is in the sandbox, its API calls need to be online Some toggle it on/off for every API call, some install a sidecar, some disable web search tools, some use a separate shell for the cmds

Principles of Robot Motion! 📖 Another weekend, another free resource to upgrade your robot skills :) Principles of Robot Motion is a book on robot motion planning, covering theory, algorithms, and real implementations. The book makes mathematical complexity accessible by connecting low-level implementation details to high-level algorithmic concepts. Topics covered are motion planning, sensor-based planning, probabilistic planning, localization, mapping, and planning for dynamic and non-holonomic systems. What stands out is that it treats motion planning not as an isolated problem but as something applicable across domains. The techniques transfer to circuit board routing, computer graphics, robot-assisted surgery, drug design, and protein folding. This reflects advances from the last 10 years in the field. Each chapter builds on previous ones, and the mathematical derivations are quite detailed. Good starting point for an aspiring roboticist! Free book on Github: github.com/yangmingustb/plan… ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com

lib25519-20260614 released: lib25519.cr.yp.to lib25519-cr-yp-to.viacache.n… Many new and improved ARM implementations from Kaushik Nath. Improved integration of verified code from s2n-bignum. Ported to more environments. CPUID tests and compiler lines are now auto-generated.

An interesting use of the improve architecture skill, is using it to analyse a newly created PRD before developing it. I’ve seen noticeably better implementations because of it @mattpocockuk

Composer 2.5 by cursor for implementations

what makes them different from the other x402 implementations tho?

Don't let software limitations define your competitive edge. Customize or build solutions to support unique business processes and achieve faster, more successful ERP implementations. #BusinessStrategy #ERP

No, we don't know. What exactly elevates him, his interpretations, and his implementations above the Sahaba and the scholars who came before him? Kindly enlighten the rest of us, who are guilty of nothing but ignorance.

🧵1/3 [Phlame Protocol v1] ◆ Implementations: • Analysis of on-chain metadata from Universal Profiles • Off-chain Bloom Score calculation • User-friendly UI/UX • Key Features (daily rewards, airtrust, LSP7/8 SBT generator & manager, mutuals back) 🔗 phlame-protocol-v1.vercel.ap…

a founder paid us $3,500 to fix a codebase last month that another AI tool built in 6 months. 400 files. all AI generated. client thought it was production ready. here's what we actually found inside: - 4 API keys hardcoded directly in the source files - no error boundaries anywhere in the app - 3 different auth implementations running in parallel - database schema with no foreign keys or indexes - 5 npm packages that don't actually exist - every component doing 4 jobs instead of 1 the app worked. it would have collapsed under 100 users. 6 months of building. $3,500 to clean up. weeks of delay before the founder could actually ship to real users. vibe coding is a speed tool, not an architecture tool. the founders who understand that distinction ship fast and clean. they use AI to execute a structure that was already decided. the ones who don't spend months rebuilding something that needed 3 weeks of proper thinking upfront. write the spec and define the schema and only then prompt.

"make a kernel" I'm afraid your design is already doomed if you think a big kernel is needed, or that classical syscalls should be the main way one exchanges information what you really want is a good L4 microkernel for the absolute basics, then build the userspace on top, and great L4 implementations exist already

You are an MPA, stop with victim mentality & talk about solutions/implementations By the way, what justice your govt. has provided to the Australian family who lost their child because of terro*ist CCD? Look at the bullet holes, not even a terro*ist would be shot at like this

A few days ago, I bought my first pair of reading glasses. The timing was interesting: they arrived almost exactly when I finished writing my book after eight months of work. Since then, I’ve heard the same comment repeatedly: “With AI, writing a book isn’t really that difficult anymore.” Maybe. But that’s not the interesting question. The interesting question is what remains valuable in a world where knowledge, content, code, analysis, presentations, and even entire books can be generated in seconds. After 28 years in technology, I’ve watched technology waves come and go: Client-Server, the Internet, Mobile, Cloud, Big Data, Digital Transformation, Blockchain, Data Science, and now Generative AI. Some changed the world. Some disappeared. AI belongs to the first category. Not because it’s another tool. Because it isn’t. And yet, after hundreds of client discussions, transformation programs, software projects, platform implementations, and AI initiatives, I keep observing the same pattern: Organizations rarely fail because they lack technology. They rarely fail because they lack strategy. Almost everybody has a strategy. Almost nobody has outcomes at scale. That question became the foundation of my book: 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗜𝘀 𝗚𝗼𝗼𝗱. 𝗘𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 𝗜𝘀 𝗕𝗲𝘁𝘁𝗲𝗿. Why Organizations Struggle to Scale AI This book is not really about AI. AI simply made the underlying problem impossible to ignore. iamyb.com/book/#store

The more I see of the new Siri, the more ChatGPT starts looking like Humane: an ephemeral market sandbar. It's a crazy thought, but if Siri can actually look up things, do basic tasks, and integrate seamlessly with your personal data...90% of the day-to-day consumer use case for standalone chatbots is gone. Why would I endure the friction of switching to a third-party app when native, system-wide functionality is available? Not to mention it's *augmented* functionality by virtue of its privileged, private access to your personal and communication data. I don't think we should underestimate what a significant impact this will have on AI-specific companies. Beyond a baseline capability threshold, chatbots are only as valuable as the data they have access to. By providing users a way to *securely* leverage their entire library of personal data, Apple effectively makes Siri the best chatbot on the market. It just instantly kills everything else. The value of an AI-specific company then essentially collapses to the top-end implementations of their models that expose their uniquely jagged cutting edges—most notably enterprise contracts and agentic coding. For a company like OpenAI which derives a significant amount of its value from ChatGPT, it's no wonder why they are pushing so hard to grow Codex. Or why they are considering this "super app" ChatGPT x Codex amalgamation. I was a bit confused by that before; why would you want to mess up such a good thing in ChatGPT by conflating it with a totally different use case? Well, it might be because they see the writing on the wall and are trying to convert as many ChatGPT users into Codex users before it's too late. They may not be "copying Anthropic" after all...this could be them fighting for survival. We still have to fully put Siri through its paces and see how its full profile stacks up against ChatGPT or other chatbots, but even if this iteration doesn't fully get there, it seems like only a matter of time until Siri and other OS-level implementations eat the chatbot market.