Videos

Running AI agents in production requires a clear separation between the non-deterministic planning done by an LLM...

Postman built an API context graph mapping 115 microservices and thousands of endpoints, and found that it...

Gemma 2's embedding model, open-sourced under Apache 2.0 with 740M parameters, enables private on-device search by...

Claude Motion Design 2.0 is being pitched as a way to build scroll-stopping websites that combine AI-generated...

Anthropic engineers cut over 80% of Claude Code's system prompt for Opus 5 and Fable 5-class models with no...

Haiku 5.5's two-tiered pricing makes it cheap for high-volume tasks under 100k tokens, but five times more expensive...

Matt Gibiec opens a conference talk by framing AI agents as tools that are replacing human work but have no nervous...

Higgsfield's AI influencer studio can re-clothe and re-scene a single generated character across six different...

Nexus is experimenting with AI agents as players in MMO RPGs to create dynamic, spontaneous experiences and solve...

Reducto's product lead claims frontier models get you only about 60% of the way on clean documents, and the real...

WorkOS built an internal platform, Atlas, with a CLI called wow that lets any employee build, deploy, and share...

Microsoft's Windows and Surface event showed a coding agent session that sent 1.6 million tokens into a model with a...

Photocraft, a free Rust-based Photoshop clone, passes 307 of 309 PSD resave tests but only has an enforced minimum...

LangChain's memory framework for agents treats traces as raw material, filtering them into durable context (facts,...

Anthropic's own documentation confirms that long conversations, where Claude must summarize older messages, consume...

Uber's Omni MCP proxy solves the problem of an agent's context being swamped by thousands of tool definitions by...

Mistral Large 4, a 1.05 trillion parameter mixture-of-experts model, will not fit in consumer hardware: its 4-bit...

Bigger context windows don't fix agent failures when tools return large data — the model forgets the middle of the...

Jev is proving most useful as a decision-model layer inside AI automations, and this video walks through the...

Claude Haiku 5.5 is priced at 10 cents per million input tokens and 50 cents per million output tokens for prompts...

Personal AI assistants can now negotiate bills, cancel subscriptions, and even control a Tesla when given API access...

Snyk's Evo continuous offensive security uses LLM agents to perform AI-driven pentesting on every code change, not...

Anthropic's Claude Haiku 5.5 is 90% cheaper than Haiku 4.5 for most prompts and shows major capability improvements,...

Factory's software factory concept proposes that engineers shift from writing code to governing AI agents that...

Bend's thesis is that the future of AI agents requires a dedicated communication layer, not repurposed human...

Google DeepMind's new Interactions API and anti-gravity agent give each agent its own managed sandbox environment,...

Fewer than a third of companies deploying AI agents in software development have scaled them to business impact, per...

C2C (cache-to-cache) passes a sharer model's KV cache directly to a receiver model via a trained fuser, skipping...

Opus 5.5 can create motion graphics videos by analyzing a reference video and adapting it to a user's app, with the...

Runlayer improved a Chromium-in-AWS-Lambda task from 36% to 100% success by letting agents distill skills from both...

An AI agent can use MCP to drive decompilers like Ghidra and IDA Pro, letting a model enumerate symbols and follow...

Grokbot will now automatically route tasks to the best AI model for each job, including Claude Opus 5.5, MidJourney,...

A single loss change, swapping MSE for RMSSE in Omniquant, recovers about 63% of the accuracy gap when quantizing a...

For enterprise document search at scale, pure grep-style file traversal fails because documents are large,...

AI companies are accidentally building banks because their billing infrastructure checks usage only after invoicing,...

Mistral Large 4, aka Lechonk, is a 1 trillion parameter open-weights model built from scratch in Europe. The name is...

Resolve AI's agentic harness for production systems is built on six pillars — model orchestration, context...

Speculative decoding can speed up LLM inference by 1.6x on structured tasks like code or JSON generation, but offers...

PlanetScale argues that the 'move fast and break things' ethos is outdated for AI-era infrastructure; instead, you...

A 300B parameter model can run on a 128GB mini PC using aggressive quantization to ~2-3 bits per weight, sparse...

Building an eval platform for AI agents is fundamentally a systems problem, not a UI or spreadsheet issue, because...

Anthropic released 13 new skills for its small business plugin, bringing the total to 44. The key conceptual...

Self-improving agents that use production traces to propose fixes and update their own evaluation data sets are...

Higgsfield's AI Influencer Studio allows creating two distinct AI characters from scratch and transferring their...

A builder is running Kimi K2, a 2.8 trillion parameter model, on 16 Nvidia DGX Spark boxes at about 21 tokens per...

Harness engineering is the emerging discipline of building the loops that turn coding agents into a software...

Holo4 is a model trained to choose between clicking, coding, and calling tools, unifying three agentic lanes that...