Category: tech > ai

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.