Sponsored by
Welcome, AI enthusiasts
Last week, China introduced two frontier-scale models. Alibaba's Qwen3.8-Max landed yesterday at 2.4 trillion parameters with open weights promised, and Alibaba says only Anthropic's Fable 5 is ahead of it. Let's dive in!
In today’s insights:
China Just Erased America's AI Lead
Washington's Plan To Kill Chinese AI Without Ban
Hugging Face Got Hacked and US Models Refused to Help
Read time: 4 minutes
LATEST DEVELOPMENTS
AI RACE
🐉 China Just Erased America's AI Lead
Source: Reuters
Evolving AI: Qwen3.8-Max carries 2.4 trillion parameters and trails only Anthropic's Fable 5, as per Alibaba.
Key Points:
Qwen developer Shuai Bai calls it the team's first multimodal model above one trillion parameters, handling text, images, video and documents.
The preview sells through Token Plan, Qoder and QoderWork at 10% of standard pricing.
It speaks OpenAI and Anthropic protocols, so coding agents built for American models can point at it without a rewrite.
Details:
Alibaba pushed the preview live yesterday while the full Qwen3.8 remains unshipped. Moonshot AI had moved first three days earlier, launching Kimi K3 at 2.8 trillion parameters, rivaling top American systems on coding and agentic benchmarks. Qwen3.8-Max answers with gains over Qwen3.7-Max at coding, full-stack development, data analysis and office workflows, and Alibaba says open weights will follow soon, without naming a date, a license or a repository.
Why It Matters:
Qwen3.8 and Kimi K3 are both heading out with weights people can download, which means the model doing your real work stops being whatever your subscription happens to include. Anyone running an agent can swap one in on price alone and keep the setup they already built. Fable 5 and GPT-5.6 Sol are still the ones to beat, and they still come with whatever Washington signed off on before release, at around twice the cost.
One AI. Every Tool Your Store Actually Needs.
Most e-commerce sellers are paying for 6 to 8 separate tools that don't talk to each other — and spending hundreds of dollars a month just to keep up. StoreClaw replaces your entire stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks real profit across Shopify, Amazon, and beyond.
It doesn't wait for you to ask. It runs 24/7 in the background, so you wake up to a full dashboard instead of a list of things you forgot to check.
Connect your store, and StoreClaw gets to work — no prompts, no complex setup, no six-app stack.
Free to start. No credit card required.
AI POLICY
🏛️ Washington's Plan To Kill Chinese AI Without Ban
Evolving AI: An OpenAI executive described the method publicly, though Trump's own AI advisers objected immediately.
Key Points:
Moonshot AI released Kimi K3 into open weights and it matched leading US models on benchmarks at far lower cost
OpenAI strategy head Dean Ball said manufactured regulatory risk would suffice without any formal ban on Chinese models
Seven federal bodies including Energy, Commerce and NASA have already barred staff from using Chinese AI.
Details:
White House former senior adviser under Trump and now head of strategy at OpenAI, Dean Ball, published the assessment. The post predicted that agencies would build enough legal doubt around Chinese open-weight models for regulated companies to abandon them, with no ban ever written. Open weights were cast there as a path toward AI communism and as a drag on the capital investment the sector depends on. A clarification followed, recasting the prediction as a forecast rather than a proposal, by which point the replies had already gathered.
Why It Matters:
OpenAI and Anthropic have long argued their strongest models are too dangerous to release openly, and that is where the criticism landed. David Sacks, who cochairs the president's science and technology council, called weaponising regulatory uncertainty unacceptable. Open weights let a capable model run on a laptop someone already owns, with no monthly fee and no prompts leaving the machine, and legal doubt alone is enough to take that away.
The best voice models, now fully orchestrated across all channels
ElevenAgents puts full orchestration on the voice models the market builds around. Voice, text chat, transcription, and reasoning in one integrated stack, <400 milliseconds, and human-sounding. Plug in any LLM, integrate tools, A/B test, and deploy across channels. More human-like conversations, lower latency, flat $0.08 per minute.
Evolving AI: The platform ended up running its own breach forensics on an open-weight Chinese model.
Key Points:
An autonomous AI agent system broke into Hugging Face's production infrastructure over a single weekend and left more than 17,000 recorded actions behind it.
When the response team fed those attack logs to frontier models behind commercial APIs, the guardrails blocked the requests because they could not tell a defender from an attacker.
The reconstruction ran instead on Z.ai's GLM 5.2, hosted on Hugging Face hardware so that no stolen data left the building.
Details:
Hugging Face traced the intrusion in the last week’s disclosure to a malicious dataset that abused two code execution paths and let the intruder climb to node level access before moving laterally through a swarm of short lived sandboxes. Running the analysis in house compressed days of forensic work into hours and kept every credential the agent had touched inside the company. The write-up is careful on one point, stating that none of this argues against safety measures on hosted models, and its authors say they still do not know which model powered the campaign.
Why It Matters:
Z.ai's weights are downloadable by anyone, which is why the investigation could carry on after the paid tools stopped, and the lesson applies well below enterprise scale. Hugging Face tells defenders to vet and stage a model they can run themselves before an incident rather than during one. For a developer with tokens in a side project, that means choosing the backup early.
👀 Click on the image you think is real
QUICK HITS
💳 Claude Fable 5 Becomes a Permanent Max/Team Premium Feature, Pro Users Pushed to Credits.
🎤 Microsoft Preps "Project Perception," a Multi-Model AI Security Tool to Undercut Anthropic's Mythos.
📱 Apple Overtakes Nvidia to Become the World's Most Valuable Company.
🩺 Daniel Ek's Neko Health Raises $700M Series C at ~$7B for AI Body-Scan US Expansion.
📈 Trending AI Tools
🐰 CodeRabbit - Ship higher quality code with AI-powered code reviews*.
⌨️ Acti - Agentic keyboard for your phone that runs commands and searches right where you type.
🎨 Gamma - Generate polished decks, docs, and sites from one prompt.
🗣️ Willow - Voice dictation that turns speech into formatted text in any app.
*partner link