Videos

1 in 20 production clinical AI notes carry serious errors that could cause significant harm, according to the...

Unsloth achieves roughly 2x speed and 70% less memory on single-GPU fine-tuning by hand-deriving matrix...

Building agents has become trivial with modern frameworks like Flu and Cloudflare, but the real challenge is...

Fine-tuning a small model for a narrow classification task built up hidden technical debt that outweighed its...

Anterior's synthetic data pipeline reverses the typical inference workflow: instead of starting with data, it starts...

X12, the decades-old standard for healthcare insurance transactions, is the ideal harness for AI agents in claims...

Ufonia, a UK-based healthcare AI company, has built a simulation framework called Matrix that uses an LLM-powered...

A 27 billion parameter model cannot be fine-tuned on a laptop GPU, primarily due to memory constraints. Full...

Todd Fisher demonstrates a guitar plugin that speaks typed or spoken text using AI tools, including text-to-speech,...

MCP now supports full JSON Schema 2020-12 for tool schemas, making tool definitions more expressive and allowing...

The Model Context Protocol (MCP) has gone stateless, removing the need for sticky sessions and session stores when...

This week four Chinese labs shipped frontier open-weight models, and Hugging Face's report shows the open-source...

Deno CEO Ryan Dahl argues that AI agents need a hard, external security boundary rather than relying on the agent...

Anthropic's own 186-page report reveals that its biological safety filters were off for 11 months, affecting 50,000...

Google's Gemini 3.7 Flash is here, and it's faster, cheaper, and more capable than its predecessors. In tests, it...

Boris Cherny, creator of Claude Code, lays out a four-step ladder from basic AI chat to full AI-native operations...

This reading group session argues that the next evolution in AI-assisted software development is 'loop engineering'...

A photo-sharing app with 400 million users quietly dropped a 280-billion-parameter agent model on HuggingFace and...

Encrypted 'thinking' blocks from Claude, GPT, and Gemini can be decoded by replaying them into the cheapest sibling...

GLM 5.3 is a 753B-parameter open-weight mixture-of-experts model that matches closed-source giants like Kimi K3 on...

GLM-5.3 is the same neural network as its predecessor, but post-training on synthetic environments turned it into...

Alibaba's Qwen 3.8 27B dense model claims to beat Opus 4.6 on several coding benchmarks while being Apache 2.0...

Exo is a new agent architecture that lets an LLM agent safely edit its own code and runtime behavior, collapsing the...

China's Z.AI released GLM 5.3, a post-training-only upgrade that found 2,436 vulnerabilities across 269 open-source...

ChatGPT just got 14x faster thanks to Cerebras custom chips, making the model no longer the bottleneck—your computer...

Qwen 3.8 27B is a huge leap over 3.6 for local AI agentic coding. In a zero-shot test, it built three playable retro...

An AI dark factory is a repository that ships its own code — you just feed it a spec and it builds, validates, and...

DeepSeek V4 Pro is currently the cheaper model by a wide margin, but on Sunday its price spikes up to 11x,...

Context-as-a-Service (CaaS) is emerging as a new category for feeding AI agents structured, up-to-date web data, but...

Four new AI models hit the market in three days — Grok 4.6, DeepSeek V4 Pro, Nemotron 3.5 Lightning, and Muse...

Nate Herk pitted Claude Code against Codex (via Cursor) to build the same Typeform clone from an identical prompt,...

The long tail of the web will never build APIs, but computer-use models that see and click like humans already work...

DeepSeek shipped a new flagship model, an open-source agent harness, and a price list that raised one rate by 1,114%...

Grok 4.6 ties OpenAI's top model on a key benchmark at a fifth of the output price, but the model is the least...

xAI dropped Grok 4.6, a point release that leapfrogs coding and knowledge-work benchmarks, making it a serious third...

16 GB VRAM just became the most common GPU config on Steam, which means the local AI tier list has shifted. Qwen 3.5...

Codex's browser agent is the best I've tried — it can automate anything in a browser or on your local machine. It...

Grok 4.6 is a serious competitor to models from OpenAI and Anthropic, especially in coding and front-end tasks. The...

Nvidia's Nemotron 3.5 Lightning is 88% not a transformer — it's a hybrid of Mamba state-space layers and...

DeepSeek V4 Pro (0813) is out of preview and it's a big leap over the undercooked preview version, but it's not the...

Improving agents is fundamentally a data mining problem: the key is to collect traces of agent behavior, then mine...

Shlok Khemani reverse-engineered the memory systems of ChatGPT, Claude, and Gemini over the past year and found that...

SpaceX bought Cursor and merged with XAI to launch Grok Bot, a $200/month AI agent platform that lets non-technical...

The real cost of using AI coding agents isn't the model — it's the harness. Harnesses like Claude Code add...

Deleting most of your Claude skills and system prompts can make the model smarter, not dumber. The creator of Claude...

Nvidia dropped Nemotron 3.5 Lightning, a 30-billion-parameter mixture-of-experts model with only 3 billion active,...

Meta's Muse Glimmer is a 30B parameter coding agent that runs in 14GB of RAM thanks to dynamic quantization, which...

Meta released Muse Glimmer, a 30B-parameter open-source model (Apache 2.0) designed to fit on a 24GB GPU, built via...

NVIDIA quietly released Nemotron Lightning, a small open-weights MoE model (30B total, 3B active) built specifically...

NVIDIA's Switchboard is an open-source routing library that sits between your agent and its models, deciding...