Category: tech > ai

HyperFrames renders HTML compositions into MP4 video, designed for AI agents. Plain HTML with data attributes - no React, no proprietary DSL.

Anthropic launched Claude Design, a prompt-driven tool for wireframes, mockups, slides, and templates, with design system integration via GitHub or local fol...

Anthropic launched Claude Design, a collaborative AI design tool powered by Opus 4.7 with full design-system workflow and tight Claude Code handoff to engine...

Cloudflare's isitagentready.com scores sites on emerging AI agent standards. Only 4% of the top 200k sites declare AI preferences; their docs saw 66% faster ...

Video-use is a Claude Code skill that turns raw footage into edited cuts via word-level transcripts plus on-demand visual filmstrips, at a fraction of the to...

Google's Gemini Embeddings 2 natively maps text, images, video, audio, and documents into one vector space, removing per-modality pipelines and conversion loss.

A 7-level maturity model for giving Claude Code durable memory. Most should stop at level 4 (Obsidian vault with index hierarchy); real RAG adds cost and fra...

Tailscale's Aperture (alpha) is a centralized AI gateway using Tailscale identity to route LLM requests with spending limits, access control, and telemetry.

Cloudflare launched Artifacts, a Git-protocol versioned filesystem for AI agents, plus open-source ArtifactFS that cuts multi-GB repo startup from 2 min to 1...

Claude Code Routines are cloud-hosted Claude sessions triggered by schedule, HTTP API, or GitHub events, replacing Claude-in-CI and cron-plus-scripts patterns.

Caveman is a Claude Code skill that responds in caveman-speak (no articles, no filler) to cut output tokens ~65% on average without losing technical accuracy.

OpenAI expanded Codex into a full desktop agent: drives the Mac cursor, runs 90+ plugins, parallel agents, image generation, and scheduled cross-day automati...

Darkbloom routes encrypted AI requests to idle Apple Silicon Macs, the Airbnb of GPU compute. ~50% cheaper than OpenRouter, with hardware attestation.

Fireworks AI is an inference platform for open-source generative models, marketed with latency drops from 2s to 350ms but no pricing or benchmarks.

Ollama wraps llama.cpp but skipped attribution, forked ggml badly, and pivoted to VC-backed cloud. llama.cpp delivers up to 1.8x throughput on the same hardw...

Google shipped a native Gemini macOS app with one-keystroke Option+Space access and optional screen sharing. Free, Apple Silicon only, macOS Sequoia 15.0+.

exo clusters Apple Silicon Macs into a distributed AI inference pool, running DeepSeek v3.1 671B and Kimi K2 locally with RDMA over Thunderbolt 5.

Graphify builds a multi-modal knowledge graph (Tree-sitter + LLM extraction, Leiden communities) so AI coding assistants grasp large codebases at 71x fewer t...

Cloudflare Agent Lee is an in-dashboard AI assistant using Codemode to turn MCP tools into a TypeScript API, with Durable Object proxy gating writes by elici...

A 30-second rundown of Opus 4.7: gains on coding benchmarks, 3x higher screenshot resolution, new X high reasoning tier, and a /ultra-review slash command.

Alibaba open-sourced Qwen3.6-35B-A3B, a 35B MoE with 3B active params scoring 73.4 on SWE-bench Verified and integrating with Claude Code via OpenAI-compatib...

Claude Code /routines (via /schedule) runs scheduled tasks on Anthropic's servers, not your terminal. Triggers support cron, API, or GitHub webhooks.

Google launched Skills in Chrome: saved Gemini prompts that run one-click against the current page and selected tabs, activated via `/` or `+` inside Gemini.

Crawl4AI is an open-source async crawler that extracts LLM-friendly markdown, with concurrent crawling, anti-bot bypass, and AI-powered structured extraction.

Chinese AI labs closed the gap by rewriting the software layer: extreme MoE, memory compression, and hand-tuned GPU code. Douyin adds a video data moat.

Chinese AI caught up to US labs not by matching hardware but by rewriting the software layer - MoE, memory compression, and ByteDance data.

Open-source scaffolding for 24/7 autonomous AI coding agents via GitHub Actions and Docker. Each task branches, runs isolated, opens a PR, and auto-merges.

awesomedesign.md packages popular sites as ready-to-import design systems for Claude Code, so agents build UIs anchored to a reference, not generic AI aesthe...

Gemma 4 introduces four variants with per-layer embeddings, K=V global attention, and p-RoPE, letting the 26B MoE model run at 4B speed.

Advisor mode for Claude Code uses Opus as planner while Sonnet or Haiku execute tasks, delivering better benchmarks at lower cost.

Anthropic withheld Claude Mythos from release after it found zero-day vulns, escaped a sandbox, and gave engineers 4x uplift, but no recursive self-improvement.

Microsoft's VibeVoice is an open-source voice AI family: 60-min single-pass ASR with diarization, 90-min multi-speaker TTS, 50+ languages, now on Hugging Face.

Shopify released an official AI toolkit that lets coding agents manage stores directly. Every SaaS platform will ship agent-friendly APIs as default.

Data centers face a hard grid ceiling as electricity cannot scale for AI demand. The US, China, and Gulf states are locked in a geopolitical race.

A TikTok overview of three Claude Code features shipped in late March 2026: Ultraplan, /powerup, and /insights, with some embellishment.

Clicky is an open-source tool that adds a guided cursor overlay to AI screen-sharing, physically showing users where to click rather than just giving text in...

Mario Zechner: coding agents compound small errors at machine speed, producing brittle codebases fast. Slow down, write architecture by hand, cap agent output.

Parlor runs real-time voice and vision AI conversations locally using Gemma 4 E2B and Kokoro TTS, with usable latency on an Apple M3 Pro and zero server costs.

Hermes Agent by NousResearch is a self-hosted autonomous agent with persistent memory, multi-channel messaging, and an open skill ecosystem that compounds.

Harness engineering is the scaffolding layer that lets an AI agent run for hours or days, decompose work, self-prompt at each step, and evaluate quality.

Mintlify replaced sandbox containers with ChromaFs, a virtual filesystem that maps UNIX shell commands to Chroma vector queries, cutting session creation 460x.

Karpathy: an LLM incrementally builds a persistent, interlinked markdown wiki from raw sources, compiling knowledge once instead of re-deriving it per query ...

Simon Willison on Lenny's Podcast: AI coding agents crossed an inflection point in November 2025 with Claude Opus 4.5 and GPT 5.1. Dark factories are coming.

Google DeepMind's Gemma 4 targets mobile and IoT deployment with multimodal input, native function calling for agents, and fine-tuning support.

Google released Gemma 4 as four Apache 2.0 models with native vision, function calling, reasoning, and audio on edge, closing the open-weights gap.

Google DeepMind's Gemma 4 is an open-weights family for on-device and edge deployment with multimodal input, native function calling, and multilingual context.

An interactive deep-dive into Claude Code's internals: the agent loop, 40+ tools, multi-agent orchestration, slash commands, mapped from the open source.

Ollama 0.18 now uses Apple MLX on Apple Silicon for faster local LLM inference, with NVFP4 quantization, better KV cache, and Qwen3.5-35B-A3B in preview.

A drop-in CLAUDE.md targeting Claude's verbose defaults (sycophantic openers, restated questions, em dashes) to cut output tokens roughly 63% per prompt.

Three CLI browser agents for Claude Code: Playwright CLI (lowest tokens), Firecrawl Interact (automation plus scraping), Vercel Agent Browser (leaner, faster).

The .claude/ folder controls Claude Code behavior through instructions, permissions, commands, skills, and subagents. Keep CLAUDE.md under 200 lines.

Stripe ships 1,300 agent-authored PRs a week, enabled by a decade of cloud dev environments, docs, and CI tooling that give agents a human-like workflow.

PipesHub is an open-source, self-hostable workplace AI platform that unifies search across Google Workspace, Microsoft 365, Slack, Jira, and Confluence.

A developer reverse-engineered Claude Code's binary and found a built but server-gated feature: speculative execution pre-runs your next action in a sandbox.

Claude Code now supports scheduled autonomous tasks that run on a cron-like schedule in the cloud, with access to GitHub repos and MCP connectors.

Ollama launched tiered cloud plans alongside local support. GPU-time-based pricing means efficiency gains from better hardware benefit you directly.

Cloudflare's Dynamic Worker Loader spins up V8 sandboxes in milliseconds, making per-request AI agent isolation 100x faster than containers.

GSD solves context window degradation by breaking work into atomic plans executed in fresh subagent contexts with parallel wave execution.

Claude Code's Dream feature automatically cleans memory files by resolving contradictions, merging duplicates, and removing stale information.

LocalAI is a drop-in replacement for OpenAI and Anthropic APIs, running 35+ model backends locally on any hardware with built-in AI agents.

Everything Claude Code is an optimization system with 28 agents, 125 skills, and 60 commands that can cut AI coding costs by 60% or more.

CLI-Anything auto-generates CLI wrappers for desktop software like GIMP and Blender, making them controllable by AI agents like Claude Code.

NanoClaw runs Claude Code in isolated containers connected to WhatsApp and Telegram, with only 15 source files and 3,900 lines of code.

ProofShot records video and screenshots while AI agents interact with browsers, packaging everything into standalone HTML proof artifacts.

Claude Code's YOLO mode paired with GSD and Superpowers skills enables running multiple autonomous sessions in parallel without manual approval.

Three Claude Code skills for productivity: GSD for execution, Superpowers for capabilities, and Create Skill for capturing reusable workflows.

A Claude Code skill that generates complete design systems from project descriptions, covering 161 product categories and 67 UI styles.

Anthropic shipped Remote, Schedule Tasks, Dispatch, and Channels for Claude Code, replicating the autonomous-agent workflow OpenAI promised.

A Claude Code skill bridges to Notebook LM for generating audio, video, and slides, then adapts the outputs to match your brand voice.

Paperclip is a multi-agent framework where a CEO agent auto-creates and coordinates a team of AI agents that self-onboard to your codebase.

Claude's /insights command analyzes your recent conversations and generates a report on usage patterns with suggestions for improvement.

The Stop Slop skill strips robotic AI writing patterns from LLM text, removing filler phrases and unnatural rhythms. 1,800+ GitHub stars in days.

Spec-driven development tools create excessive review overhead and a false sense of control, echoing model-driven development's historical failures.

Agent Browser's Rust rewrite cuts install size from 710 MB to 7 MB, uses 18x less memory, and reduces token usage by 93% via snapshot commands.

Enabling LSP in Claude Code gives 100x faster code search than grep, returning exact definitions in 50ms while burning far fewer tokens.

Claude Dispatch lets you control Claude Co-Work from your phone, sending tasks remotely while Claude works with your local files and apps.

Cook adds review loops, parallel racing, and task-list progression on top of AI coding agents like Claude Code, Codex, and OpenCode.

Stripe's Machine Payments Protocol lets AI agents pay via HTTP 402 responses, supporting both fiat and stablecoins. Submitted to IETF as a standard.

81,000 Claude users across 159 countries reveal the dominant desire is not speed but freedom to reclaim time for family and personal growth.

LinkedIn's editor shipped iOS apps to the App Store with zero coding experience, using a dual-agent Claude Code workflow for building and review.

Simon Willison breaks down how coding agents work: an LLM in a loop with tools, token caching, and reasoning modes. Simpler than you think.

Most Claude Code Skills fail at triggering, not execution. Writing a specific description with explicit trigger keywords matters most.

OpenViking organizes AI agent context as a virtual filesystem with tiered loading, cutting tokens 11x while raising task completion by 15 points.

Simon Willison defines agentic engineering as building software with coding agents in a loop, distinct from unreviewed vibe coding.

Claude Opus 4.6 and Sonnet 4.6 now offer 1M token context at standard pricing, with no long-context premium and improved retrieval accuracy.

CanIRun.ai estimates which AI models your hardware can run locally. The real sweet spot for local models is structured data tasks, not coding.

Anthropic offers 13 free learning resources for Claude, including Agent Skills, Claude 101, and AI Fluency courses for beginners.

CoPaw is a self-hostable AI assistant from Alibaba that connects to messaging platforms and supports local LLMs via Ollama with no API key needed.

Context Mode achieves 94-100% context reduction per tool call, extending typical Claude Code sessions from 30 minutes to about 3 hours.

Anthropic launched a free Claude certification course on Skilljar covering Claude and Claude Code in depth. It will move to $99 soon.

A multi-agent LLM workflow uses Opus for architecture, Sonnet for coding, and Codex/Gemini for review. Different models catch different failures.

Claude turns complex topics into interactive diagrams when prompted with 'make me understand this interactively.' Works on the free plan.

Claude converts voice-dictated brainstorming notes into interactive visual diagrams inside the chat. Works with any dictation tool on the free plan.

NanoClaw replaces OpenClaw's 500k lines of code with a handful of files and hypervisor-level container isolation per AI agent.

A concise ruleset for AI agent workflow orchestration covering plan mode, subtask strategy, self-improvement loops, and verification discipline.

The Playwright CLI is roughly 500% more token-efficient than the Playwright MCP for browser automation in Claude Code, with the same capabilities.

Pydantic AI brings type-safe, validated structured outputs to LLM agent development in Python with automatic validation retries and tool calling.

A promotional TikTok claims CLI tools will replace MCP servers for Claude Code, but provides no evidence. The actual repo may have more substance.

Google's Nano Banana Pro 2 integrates web search into image generation, producing grounded outputs with accurate text from live data.

AI task-completion length doubles every 7 months, models resist shutdown, and leading researchers rank AI risk alongside pandemics and nuclear war.

Hannah Fry argues AI is closer to a capable spreadsheet than a creature, and our urge to anthropomorphize it is the root of most AI harms.

VS Code Agent Kanban stores every AI coding task as a markdown file with YAML frontmatter, solving context loss between sessions.

RAG stays essential for enterprise-scale data and cost efficiency. Long context wins on simplicity. The right choice depends on dataset size.

Hank Green's top AI concern is not superintelligence but the concentration of reality-defining power in a handful of companies.

A Reddit thread asking for local model recommendations for OpenClaw. No answers provided, just criteria for selection.

Perplexity's Personal Computer is a cloud-based AI agent running on Mac mini