标签：#us

Elon Musk's tech company released the newest version of Grok on Wednesday, promising a cheaper, more efficient alternative to other powerful AI models.

Have you ever wondered how AI engineers fine-tune and align large language models? Under the hood, they run Supervised Fine-Tuning (SFT), optimize parameters using direct preference gradients (DPO), filter out low-quality pre-training corpuses (Pruning), and mitigate catastrophic drifts. To help you visualize how LLM alignment and parameter optimization work in a highly strategic way, I built a cyberpunk card battler inspired by Gwent: 🤖 Epoch Duel: Cyberpunk LLM Alignment Battle Play in Fullscreen Mode (if the embed sizing is tight) 🛠️ Tune Your Model Parameters Your mission as an alignment engineer is to play optimizer cards to outscore the adversarial baseline AI across 3 training Epochs: ⚙️ Logic & Coding: Run SFT code snippets, compile theorem provers, and deploy Python scripts to build your coding benchmark scores. 📖 Language & Speech: Train on multilingual datasets and summarization corpuses to maximize reading comprehension. 🛡️ Safety & Alignment: Implement red-team safeguards, configure RLHF preference pairs, and run DPO tuning to protect your model's outputs. ⚡ regularizers & Drifts: Deploy Regularization cards like Gradient Clipping (Scorch) and Model Pruning to destroy anomalies, or exploit Anomalous Drifts to collapse the AI's rows. 🧬 Playable ML Concepts Explained Here is how the card battle mechanics map to production machine learning pipelines: 1. ✂️ Model Pruning (Weight Compression) In-Game: Playing the Model Pruning card triggers a glitchy dissolution animation that purges the lowest-value card from the targeted board row, cleaning up noise. 💾 The Real-World Counterpart Model Pruning removes unimportant weights (often those closest to zero) from a trained neural network. It shrinks the memory footprint of the model, allowing it to run faster on edge devices. ⚠️ How it affects LLMs By stripping out low-impact weights, pruning compresses models by 30-50% with minimal loss in benchmark accuracy, making deployment significantly cheaper. 2. 🔀 DPO vs RL

AI coding agents are becoming good enough to touch real codebases. They can refactor files, write tests, change architecture, move logic around, and sometimes modify more code in ten minutes than a human would in an afternoon. That is powerful. But it creates a new debugging problem. Git can tell you what changed . When an AI agent was involved, you often need to know something deeper: Why did this change happen? Which prompt caused this line? Which model produced it? What files did the agent read before writing it? What later changes depended on this agent action? That is the problem I wanted to solve with Causari . Causari is a local CLI for intent-addressable code . It records AI agent actions as causal events: prompts, models, reads, writes, diffs, reasoning, cost, and relationships between actions. The goal is simple: Git tracks bytes. Causari tracks intent and causality. Repository: https://github.com/croviatrust/causari Website: https://causari.dev The problem When a human developer changes code, there is usually some context. A commit message. A pull request. A ticket. A discussion. A design decision. With AI coding agents, the workflow is different. You ask something like this: Refactor the auth flow and add JWT refresh logic. The agent reads files, makes assumptions, writes code, maybe fixes tests, maybe changes something unrelated, then moves on. At the end, you have a diff. But the diff does not tell the full story. A suspicious line appears in auth.ts . Git can show when the line appeared. But Git cannot answer: which prompt produced this exact line? what completion did it come from? did the agent read the right files first? was this part of the original request or an accidental side effect? if I revert this, what downstream work am I also undoing? That gap becomes bigger as agents become more autonomous. The more work agents do, the more we need provenance. Not only code provenance. Intent provenance. The idea: intent-addressable code Causari treats an

But the devices could "easily be reflashed."

GitHub热门项目 | RuVector is a High Performance, Real-Time, Self-Learning Ai, Vector GNN, Memory DB built in Rust. | Stars: 4,341 | 8 stars today | 语言: Rust

With funding from tennis star and angel investor Venus Williams, the French app WeWard says that it increases walking time by almost 25%.

Quantum mechanics gave us the transistor before we understood it. The same pattern is happening with AI right now — and the builders who recognize this will define what comes next. The argument that never ended — and the lab that didn't care In 1927, the greatest minds in physics gathered in Brussels for the Solvay Conference. Albert Einstein, Niels Bohr, Werner Heisenberg, Erwin Schrödinger, Max Planck, Marie Curie — twenty-nine of the most brilliant humans who ever lived, in one room. They were arguing about quantum mechanics. Specifically: what does it mean for a particle to exist in multiple states simultaneously until observed? Does reality require an observer? Is the universe fundamentally probabilistic? Is God playing dice? Einstein said no. Bohr said yes. Neither convinced the other. That argument never fully resolved. Nearly a century later, physicists still debate the interpretation of quantum mechanics — the Copenhagen Interpretation, Many Worlds, Pilot Wave theory. We have not settled it. Meanwhile, in 1947 — twenty years after the Solvay Conference — three engineers at Bell Labs in New Jersey quietly invented the transistor. William Shockley, John Bardeen, and Walter Brattain did not wait for the philosophical debate to conclude. They did not need to understand why quantum tunneling worked at a fundamental level. They understood it well enough to build something with it. That transistor became the foundation of every computer, every smartphone, every server, every piece of digital infrastructure that exists today. We built the entire digital civilization on something we still don't fully understand. Not despite the uncertainty. With it. The pattern repeating right now Across the internet in 2025 and 2026, a remarkably similar argument is happening. Will AI take all the jobs? Is it conscious? Does it hallucinate too much to be trusted? Are we building something we cannot control? Should we slow down? Should we stop? These are not trivial questions. The r

For more than a decade, one question has loomed over the race to build autonomous vehicles: Are cameras alone enough to safely replace human drivers, or do truly driverless cars need additional, overlapping sensors like lidar and radar to navigate the world reliably? Tesla has bet billions of dollars that artificial intelligence and cameras are […]

Verity Harding tells WIRED that the US government’s nationalistic attitude toward AI is evidence that a worst-case scenario is taking shape.

A key delay from a state agency means robotaxi rides in the company’s new Ojai vehicle might be free for a few more months.

ZML, a hot French AI startup endorsed by Turing Award winner Yann LeCun, has now released ZML/LLMD, software that could make running AI less costly.

Intro AI Avatar is a completely free app that lets your VRoid (VRM) 3D avatar animate in...

Is it just me, or has anyone noticed that articles on dev.to don't get as many reads/views as they...

Last week I sat in my parked car for fifteen minutes before going inside. Not because of...

Here is an uncomfortable one: you do not own your reading list. You rent it. Every "follow" button you have pressed in the last decade put your reading relationship inside a company's database, where it can be ranked, throttled, or ended the day the business model changes. You did not sign anything. You just stopped owning it. It was not always like this. Feeds were the quiet machinery that kept the web interoperable. RSS and Atom meant a site, a reader, and a robot could all agree on the same stream without asking anyone's permission. You published once, and anything could read it: whatever app, whatever order, no algorithm in the middle. Then it eroded. Plenty of sites ship no feed at all now, and "follow us" quietly became "create an account on someone else's platform." The reason is not mysterious. Platforms had every incentive to close the loop, because a feed lets you leave, and an account does not. So the industry swapped "here is my stream, read it however you like" for "log in to see updates," and a generation of sites simply stopped publishing feeds, because the platform was where the audience was. That is the trade you made without noticing. The open format that asked nothing of you got replaced by a login that asks for everything. Your reading list used to live in your reader and survive a company changing its mind, its ranking, or its whole business. Now it lives in their database and survives exactly as long as they allow. Getting it back is not nostalgia. It is infrastructure for independence: tooling that treats feeds as a first-class citizen, aggregates the sources you actually choose, and keeps that stream under your control instead of a platform's. The full case for why this is worth fixing, and what feed-first tooling looks like, is here: https://mederic.me/blog/open-web-feeds So, honestly: how many of the people and sites you follow could you still read tomorrow if the platform in the middle disappeared tonight?

If you're building something ambitious, this is a fast track to the people who can move your startup forward.

The new image-generating model has numerous use cases, including advertising, decorating and creator-based opportunities.

Two weeks ago one of my outdoor cats bit me. She's fine — healthy, pregnant, and deeply offended that I picked her up, but she needed flea medicine and I needed to confirm the pregnancy. (If anyone wants a kitten, I know a grumpy lady who has some.) My pinky swelled up, and typing went from "mildly error-prone" to "not happening." So I dictated this post. If you've ever looked at raw voice transcription, you know what that produces: one giant unpunctuated block with half the words wrong. My transcript literally claims "AI needed to put flea medicine on her." It was me. That's the kind of thing the AI is cleaning up. The ideas are mine. The argument is mine. The punctuation and clarification belongs to the machine, because the machine is better at punctuation than a transcript is. By the rules of the current discourse, you're now supposed to stop reading. That's the game, right? "Not reading this if it's AI-generated." "It has em dashes — slop." Let's deal with the em dash first, since it's apparently forensic evidence now. You can type one. Shift-Option-hyphen on a Mac. Windows-Shift-hypen on Windows. Writers were littering pages with them for a century before the first transformer shipped. Its little brother the en dash is everywhere too, and nobody has ever accused an en dash of being a robot. The em dash gets singled out for exactly one reason: it's a fast, cheap way to judge a piece of writing without engaging with a single idea in it. Zero effort, instant superiority. Remember that phrase — zero effort. It's coming back. Because real AI slop absolutely exists. Someone fires off one prompt, ships whatever falls out, never reads it, then farms for stars and upvotes. That's slop — not because a model was involved, but because no human was. Effort is the variable. The tool never was. Here's what the other end of the spectrum looks like. Hundreds of hours on a single project. I decide the architecture, the language, how it compiles, how it deploys. I fork the output

Have you ever wondered how vector databases like Pinecone, Milvus, Qdrant, or pgvector search through billions of high-dimensional documents in milliseconds? Under the hood, they map semantic concepts into dense numerical vectors, calculate multidimensional cosine similarity angles, and traverse proximity graphs to locate nearest neighbors without scanning the entire database. To help you visualize how vector databases and embeddings actually operate, I built a retro-vector arcade game: 🛰️ Vector Strike: Database Defender Play in Fullscreen Mode (if the embed sizing is tight) 🛠️ Choose Your Database Optimizations Your mission as a Vector Database (VDB) administrator is to configure your query settings and index structures to defend your index nodes: 📏 Similarity Threshold (τ): Tweak the match threshold slider. High thresholds require near-identical semantic matches but protect your index, whereas lower thresholds act like a splash-damage laser but risk matching incorrect clusters. 🪐 Embedding Dimensions (2D $\rightarrow$ 8D $\rightarrow$ 32D): Higher dimensions isolate categories and guarantee precise hits. Lowering dimensions collapses the projection space, causing spatial overlap that results in false deflections and friendly-fire query failures. ⚡ Proximity Indexing (Flat Scan $\rightarrow$ HNSW Graph): Flat Scan: Runs a brute-force linear search over all targets. It causes computation latency spikes as more query objects arrive. HNSW (Hierarchical Navigable Small World): Dynamically builds proximity links between adjacent node targets. The turret traverses vectors along the nearest-neighbor graph, snap-locking onto targets with zero lookup latency. 🧬 Playable ML Concepts Explained Here is how the arcade mechanics map to production vector databases: 1. 🔀 Multidimensional Projections (Dimension collapse) In-Game: You can toggle between 2D, 8D, and 32D space. In 32D space, the categories are cleanly separated. In 2D space, the database collapses, and you'll find sp

Joshua Achiam spent nearly nine years at OpenAI researching AI safety and made a memorable appearance in the Musk v. Altman trial.