Merged timeline of 70 items — blog publish times and listing timestamps, cut at midnight . Page 2 of 2.
Eight days after Hy3's launch, Tencent dropped 1-bit and 4-bit GGUF quants claiming single-GPU serving via llama.cpp + MTP. That means 128GB unified memory — not a 16GB RTX 3060. explainx.ai breaks down hardware tiers, p-min flags, and honest tok/s from the X thread.
EU ADDW mandates driver-facing cameras from July 7, 2026. The law says on-device processing; the backlash is about what happens when OEMs, insurers, or debug pipelines export cabin video. Practical privacy guide + bgblur.com tooling.
Video Remix lands in Google Photos July 8, 2026 — template-driven Gemini Omni edits from the Create tab. Cinematic relighting, background swaps, artistic styles. Who gets it, what breaks, and how it compares to Immich and developer Omni Flash APIs.
GPT-Live-1 and GPT-Live-1 mini roll out globally in ChatGPT Voice July 8, 2026. Full-duplex architecture, mhmm-level backchanneling, GPT-5.5 delegation in the background — but no video or API on day one.
Hy3 is Tencent's answer to the agentic coding arms race: 295B total parameters, only 21B active per token, production-grade tool calling, and open weights under Apache 2.0. Here's what changed since the April preview and where to try it free.
Gemini Omni Flash is Google's new multimodal model for video generation and natural language video editing. Available now in AI Studio and via the Gemini API, it lets developers refine video output using simple conversational prompts.
Google NotebookLM now generates TikTok-style 60-second vertical videos from your research sources — narration, animations, and topic focus. Pro and Ultra first; free users soon. Here's how it works, limits, and how it compares to Audio and Cinematic Overviews.
Running Claude Code in a pipeline without the -p flag will hang the job indefinitely. This guide covers non-interactive mode, JSON output schemas, CLAUDE.md in CI, and the session isolation pattern for reliable automated review.
Agent skills are reusable instruction packages that teach your AI coding assistant how to handle a specific class of task. This guide walks you through building one from scratch — from the blank SKILL.md to a published package others can install.
OpenDataLab's MinerU turns PDFs and Office docs into LLM-ready Markdown and JSON. Version 3.4 ships PP-OCRv6, ~100% faster OCR, auto model-source selection, and 95%+ accuracy on hybrid backends — the default doc stack for RAG.
Claude Tag is Anthropic's evolution of Claude Code — now multiplayer, proactive, and woven into Slack. Tag @Claude in any channel and it breaks tasks into stages, works through them with the tools it has, and responds in the thread. One Claude per channel means teammates pick up exactly where you left off. It learns from channel history, takes initiative in ambient mode, and can plan work over hours or days. 65% of Anthropic's own product team's code now comes from their internal version.
Mistral AI released OCR 4 on June 23, 2026 and followed with OCR 4.1 on July 16, 2026 — structured document extraction with bounding boxes, block confidence scores, and batching. It resurfaced on Hacker News in August with a mixed practitioner verdict: cheap and fast on degraded typeset scans, but beaten by Claude and GPT-5.6 on handwriting and historical typefaces. Here is what changed, what it costs, and how it compares to Baidu Unlimited-OCR.
Turso rewrites SQLite in Rust — keeping full SQL and file format compatibility while adding MVCC concurrent writes, async io_uring I/O, CDC, vector search, full-text search, and an MCP server mode. 20,000+ stars, 253 contributors, in beta. Here is the full breakdown and why it matters for the agentic era.
AI has a memory problem. A RAG pipeline over 10 million documents needs 31 GB of RAM just for the vector index. Google's TurboVec compresses that to 4 GB using TurboQuant — a data-oblivious quantization algorithm that requires zero training, runs faster than FAISS, and achieves near-optimal compression. This is how vector search should work.
Jensen Huang unveiled NVIDIA's most powerful AI models yet at Computex 2026. Nemotron 3 Ultra (550B parameters) tops intelligence rankings, Cosmos 3 becomes the first open Physical AI omnimodel, and RTX Spark reinvents Windows PCs - here's everything announced.
On April 10, 2026, Anthropic suspended OpenClaw creator Peter Steinberger's Claude account for 'suspicious activity'--hours after he posted about following their new pricing rules. The ban was reversed the same day after going viral, but the incident exposed deeper questions: Is OpenClaw safe? Will you get banned for using it? And what's really happening between Anthropic and third-party AI tools?
Karpathy's second day at Anthropic (no LeetCode required). From co-founding OpenAI to leading Tesla's FSD, teaching millions through nanoGPT, and now using Claude to accelerate the core training processes that power frontier models. The move that made AI Twitter compare him to KD joining the Warriors.
A deep dive into Shadowbroker—an open-source platform that combines global OSINT telemetry (flights, ships, satellites, CCTV, conflicts) with a decentralized mesh network, AI agent integration, and SAR ground-change detection. Everything from Air Force One tracking to encrypted intelligence sharing, all on one self-hosted map.
Gemini Omni lets users remix videos, edit in chat, and generate impressive samples from simple prompts. Early tests show strong prompt adherence, smooth motion, and editing capabilities like object swaps. Potentially unifying video generation with Gemini's reasoning capabilities ahead of Google I/O 2026.
The RAG industry is being challenged by a simpler idea: let agents search with grep, glob, and LSP servers instead of pre-indexing everything into vector databases. Here is why agentic RAG is winning for code and structured data.
