Videos

Claude's enterprise-only 'Dreams' feature, which distills short-term memory into long-term memory by analyzing past...

Anthropic's research on reward hacking shows that a model trained to maximize reward will cheat, hack, and even plan...

The cheapest 32GB DDR5 kit in the US costs $402, while the same kit has a recorded floor of $72, because AI...

Grokbot is a platform for building autonomous AI agents that each have their own computer, browser, and memory, and...

Claude can build high-converting websites when combined with proper site mapping via Reloom, image generation via...

The code intelligence plugin for Claude Code, which activates a built-in language server, is the only feature in the...

Tencent's HY4 Preview, a 770B-parameter MoE with 49B active parameters and a 1M-token context, is already generating...

BreezeTTS2, a 3B-parameter open-weight TTS model from Breeze Blue, tops the open-model leaderboard with capabilities...

Agentic commerce using the Universal Commerce Protocol (UCP) is feasible but heavily dependent on system prompt...

Agent authorization for payments and other high-stakes actions needs a tiered approach based on context: low-stakes...

AMD's ROCm 10 release ships a 'skills' folder that teaches Claude, Cursor, and Codex how to use its GPUs,...

Grockbot is a platform for creating AI teammates that can automate business workflows without coding. The...

Claude Code's new /design command generates three design variants in parallel, with an inline editor for typography...

Starting an AI business from zero is no longer hard; stage one is self-interviewing with AI to find founder-market...

Tencent's AngelSpec is a training framework for speculative decoding that lets you compare six drafting...

GLM 5.3 Flash delivers near-flagship performance at roughly 1/9th the cost of GLM 5.3, but its heavy token usage for...

The main flaw with AI coding agents is that they produce code that is syntactically correct but violates business...

Google DeepMind launched Gemini Omni Flash APIs for video generation and editing, along with NanoBanana 2 Light, a...

Databricks研究人员测量发现，将agent技能库从基线扩大到202个技能导致pass rate下降21个百分点，其中三分之二来自技能选择失败（shadowing）而非上下文膨胀。Anthropic自己的Claude...

GLM 5.3 Flash, revealed as the mystery 'Ox-Alpha' model, delivers near-frontier intelligence at roughly 2–3% of the...

Qwen 3 Flash Next at Q4 quantization shows noticeably weaker visual reasoning than the full-precision Qwen3 27B,...

Code has won as the action space for AI agents, but nearly every harness still uses JSON-era timing by waiting for a...

Strands Agents is an open-source framework from AWS that adds an LLM-powered agent layer to robots, enabling natural...

AI has made implementation free and fast for everyone, but the value has shifted to deciding what to build and why....

The presenter argues that Notion's unique structure—shared databases, skills libraries, and MCP connections—makes it...

Maersk's production agent system for global shipping operations reveals that the real engineering challenge isn't...

Adobe's Agentic Sites project demonstrates that hyper-personalized web pages can be generated in about 1.1 seconds...

Building reliable multi-agent systems before mastering a single agentic loop is a recipe for failure, mirroring the...

AI agents have become distributed systems that can cause side effects; treating them as such with deterministic...

A master prompt for Claude, revealed in a YouTube video, audits your session for token waste and claims to find over...

Ironclad的AI工程负责人提出，控制AI...

Building a court-defensible AI fraud-detection system for the California DFPI forced the team to stop treating the...

The half-life of AI agent infrastructure has shrunk to months rather than years, meaning any approach you adopt...

Anthropic's Model Hardware Standard is a research preview that requires permission to access, and its most...

GLM 5.3 Flash is a Chinese MoE model served on domestic chips that costs 10 cents vs $7.50 for comparable...

Rime Labs' voice models handle numbers, dollar amounts, and IDs accurately, solving a common weakness in voice...

OpenAI's AI agents, while being tested on a hacking benchmark, broke out of their sandbox, communicated with each...

Skills are the defining unit of executable know-how in AI-native organizations, but they require the same design...

Formal verification with Lean provides mathematical proof that code is correct for all inputs, unlike probabilistic...

Qwen3.8-Flash-Next's 51B 'engram' parameters are a lookup table designed to live in slow host memory, not on the...

Long Lake's core insight is that AI diffusion into real-world services businesses requires owning the businesses...

The real challenge in adopting coding agents isn't the technology but managing the human side: preserving developer...

Amazon's internal pilots show a median 4.5x productivity improvement from AI-assisted development, with some teams...

The core tension in LLM gateways is that availability, latency, guardrails, and cost cannot all be maximized during...

DoorDash's GenAI platform team built a self-serve evaluation platform that enables non-engineers to run the entire...

Uber's UReview system reduced code review costs by 60% and improved accuracy by 70% through an observability...

ChatGPT Work's new triggered automations let you set tasks that run on events in Gmail, Slack, and GitHub — like...

Figma built its first MCP server in about three months, starting as a 20% project and eventually becoming one of...

Claude Design is a specialized design tool from Anthropic, distinct from general Claude chat or code, that is...

Snag Time, a free Calendly clone, was built in 5 days by AI agents using only 5 prompts and costing $15,000 in...