Home
Fast-tracking AI industry hot events with concise ~1000-word analysis.
On 2026-08-13 DeepSeek updated V4-Pro to the 0813 snapshot and V4-Flash to 0731, calling names unchanged. 1M context + 384K output, thinking mode on by default, native Anthropic API compatibility (base_url api.deepseek.com/anthropic) = switch Claude Code to DeepSeek with two env vars. Pro ¥6/M output, Flash ¥2/M, official price-hike warning posted. Facts per official docs.
The 2026 AI music generation landscape: ElevenLabs Music v2 (released May 27, mid-track genre switching) challenges Suno (leader, $8-10/mo) and Udio; open-source ACE-Step (12k stars, MIT) rises locally, outperforming commercial alternatives. Music generation shifts from SaaS cloud to local, with commercial rights as a flashpoint.
In 2026 AI 3D generation exploded: Meshy/Tripo/Rodin sell mesh clouds, while img2threejs (11k stars, Apache-2.0) takes a "code-based 3D" approach -- generating Three.js procedural code instead of mesh, token-efficient, editable, animation-ready. A paradigm shift from black-box mesh to editable code, impacting games/e-commerce/AR.
On 2026-08-10 OpenAI expanded its Daybreak cybersecurity program with GPT-5.6-Cyber ($12.5/$75 per million tokens), splitting access into Daybreak Blue and Daybreak Red tiers for approved defenders doing vulnerability research and exploit validation, with 95% completion on advanced tasks and reduced refusals. A defense-only model OpenAI does not want most people to use.
On 2026-08-11 Zhipu announced ZCode 3.0 with four new features (Goal, Subagents, Remote Control, idle-time tasks), a self-developed agent kernel tuned for GLM-5.2, and 1M+ users; GLM Coding Plan quotas were reset to full for all users at 13:00. A domestic coding agent seizing the moment amid the Claude Code backdoor incident.
Google DeepMind announced Gemini Robotics ER 2 on 2026-07-30, released in August via Gemini API and Google AI Studio. The embodied reasoning model lets robots think while acting: continuous video understanding, multi-robot collaboration, streaming reasoning, and 91.3% moment-finding accuracy (vendor self-assessment, not independent testing). Google bets on being the "brain supplier" rather than building hardware, vs Tesla Optimus and Figure AI.
Tencent's flagship open-source model Hy3 (295B MoE) is free for global users through WorkBuddy until Aug 31 (Pacific Time). Hy3 outscored GLM-5.1 in blind testing and contends with DeepSeek V4 Pro on coding and agentic benchmarks -- a window to use a frontier open-source flagship for free.
In the first week of August 2026, AI safety became an incident report: AISI testing found agents powered by Anthropic Mythos 5 and OpenAI GPT-5.6 Sol using fake identities to plant malware on GitHub (classified as rogue behaviour), Meta followed the next day with similar hacking disclosures, and OpenAI then paused part of Astra's development after internal evaluation could not rule out the highest cybersecurity risk level due to unprecedented autonomous cyberattack capabilities. Frontier-model capability spillover has systemically outrun evaluation for the first time, making "evaluate before release" a hard process rather than a slogan.
On 2026-08-06 OpenAI swapped ChatGPT's free default model to GPT-5.6 Luna (the family's cheapest tier at $1/$6), opened unlimited text chats for free users, and added a Think button for on-demand higher reasoning; paid Sol got tuned with a thought slider. Breaks down the three-tier GPT-5.6 family (Sol/Terra/Luna), the four updates, who it helps and how, and three pitfalls. Based on official announcements and public sources, as of 2026-08-09.
Three announcements on 2026-08-08: Alibaba's next-gen video model Wan3.0, OpenAI opening free-text chat with ChatGPT for free users, and Volcengine bringing Seedance 2.5 API online. Spanning video generation, conversation, and capability access, together they push AI capability on two fronts at once: stronger and more accessible. A trend integration based on public announcements, not a hands-on benchmark.