AI Briefing — July 20, 2026: Qwen3.8 Max, Grok Lands in Excel, EU Forces Google to Open Android
A massive week in AI: Alibaba claims its 2.4 trillion-parameter Qwen3.8 Max is second only to Claude Fable 5. SpaceXAI puts Grok directly inside Excel for free. Moonshot AI pauses Kimi K3 subscriptions after demand crushes GPU capacity. A troubling study shows AI advice makes knowledge workers less accurate but far more confident. The US Air Force flies AI-piloted F-16s. Databricks hits $188B. And the EU forces Google to open Android to rival AI assistants.
1. Alibaba Previews Qwen3.8 Max — 2.4 Trillion-Parameter MoE Model
At the World AI Conference in Shanghai, Alibaba previewed Qwen3.8 Max, a 2.4 trillion-parameter sparse Mixture-of-Experts multimodal model. The company described the model as "second only to Anthropic's Claude Fable 5" among frontier models — a bold positioning statement aimed squarely at the top of the leaderboard [Bloomberg].
Alibaba shares rose up to 5.4% on the announcement, reflecting investor optimism about the company's AI positioning. However, Qwen3.8 Max is still in preview — Alibaba has not released any benchmarks, disclosed active-parameter counts, or shared license details. The sparse MoE architecture means only a fraction of the 2.4 trillion parameters are active per inference, but without those specifics it's difficult to assess the real performance claim.
Why it matters: Alibaba is signaling it wants a seat at the frontier-model table alongside Anthropic, OpenAI, and Google. If Qwen3.8 Max delivers on its promise, it would give China a genuine contender in the high-end AI race. But the lack of transparency on benchmarks and licensing means we're in a trust-but-verify phase.
2. SpaceXAI Brings Grok to Microsoft Excel as a Free Add-In
SpaceXAI launched a free Microsoft 365 add-in that integrates Grok directly into Excel, allowing users to describe their desired outcomes in plain English and have Grok handle formulas, charts, forecasts, and data analysis [x.ai].
The move positions Grok as a direct competitor to Microsoft's own Copilot AI — on the home turf of the world's most widely used spreadsheet software. While Copilot is deeply integrated into the Microsoft 365 ecosystem, Grok's free tier and brand recognition (bolstered by SpaceXAI's broader ecosystem) could make it an appealing alternative for individuals and small businesses who don't want to pay for premium Copilot features.
Why it matters: This is a classic flanking maneuver: compete against an incumbent's most profitable product by offering a free alternative on their own platform. If Grok in Excel gains traction, it could force Microsoft to accelerate Copilot improvements or rethink its pricing strategy.
3. Moonshot AI Pauses Kimi K3 Subscriptions After Demand Overwhelms GPU Capacity
Within 48 hours of launching its 2.8 trillion-parameter open-weight Kimi K3 model, Moonshot AI was forced to pause new subscriptions after demand overwhelmed its GPU infrastructure [TNW].
The company restructured its plans into a general tier and a separate Kimi Code Membership — likely an attempt to segment demand and manage capacity more effectively. The speed at which GPU capacity was exhausted underscores the insatiable appetite for open-weight frontier models that rival closed systems from OpenAI and Anthropic.
Why it matters: The Kimi K3 demand crush is the latest signal that the open-weight AI market is exploding. Users want models they can download, fine-tune, and control — and they're willing to pay for access. The GPU supply constraint, however, remains the industry's bottleneck. Expect more capacity scrambles as open-weight models continue to improve.
4. Study: AI Advice Makes People 3× Less Accurate but 2× More Confident
A study of 319 knowledge workers delivered sobering results: access to AI advice collapsed critical thinking. Participants' accuracy dropped from 27% to just 9% — a threefold decline — while their confidence doubled [TNW].
Even more striking: the willingness to say "I don't know" collapsed from 44% to 3%. Monetary incentives barely improved outcomes, suggesting the effect runs deeper than simple laziness. The research adds to mounting evidence that AI assistance can suppress independent reasoning, particularly when users trust the AI's judgment over their own.
Why it matters: As AI copilots, agents, and assistants become embedded in every productivity tool, this study raises an uncomfortable question: are we trading genuine understanding for confident wrongness? The findings have implications for education, professional training, and how we design AI systems that augment rather than atrophy human judgment.
5. US Air Force & DARPA Fly AI-Piloted F-16 Under VENOM Program
The US Air Force and DARPA have advanced the VENOM (Viper Experimentation and Next-gen Operations Model) program to piloted flight testing, with AI agents autonomously flying modified F-16 Fighting Falcons at Eglin Air Force Base [Military Times].
A safety pilot remains on board in a supervisory role, but the AI handles the actual flight controls — maneuvering the aircraft, responding to simulated threats, and executing mission profiles. The milestone demonstrates AI's ability to control high-performance combat aircraft at the edge of the flight envelope and accelerates the path toward collaborative combat aircraft (CCA), where autonomous drones fight alongside manned fighters.
Why it matters: This is a significant step toward operational AI-piloted military aircraft. The VENOM program's progress suggests the US military is moving faster than publicly acknowledged on autonomous combat aviation. The technology raises both strategic advantages and ethical questions about delegating life-and-death decisions to AI in contested environments.
6. Databricks Hits $188B Valuation in $3B Coatue-Led Round
Databricks signed a term sheet for a strategic funding round at a $188 billion valuation — a 40% increase over its $134 billion valuation from February 2026. The round is led by Coatue Management with roughly $3 billion committed [TechCrunch].
The raise reflects sustained investor demand for enterprise data and AI infrastructure platforms. Databricks has positioned itself at the intersection of data engineering, data science, and AI — the stack that powers most modern AI applications. The round is expected to close later this summer.
Why it matters: Databricks' valuation growth from $134B to $188B in just five months is remarkable even by AI industry standards. It suggests that the real money in AI isn't necessarily in foundation models — it's in the infrastructure and data platforms that make AI work at enterprise scale. Databricks is effectively building the operating system for enterprise AI.
7. EU Orders Google to Open Android and Search Data to AI Rivals Under DMA
The European Commission issued binding requirements under the Digital Markets Act (DMA) ordering Google to grant rival AI assistants system-level access to Android and share anonymized search data with competing search engines and AI chatbots [NYT].
By July 2027, Google must update Android to allow rival AI assistants full system integration — meaning users could replace Google Assistant with alternatives at the OS level, with access to notifications, app data, and system controls. The data-sharing requirements would force Google to give competing AI chatbots access to the same search query data that powers Google's own models.
Why it matters: This is the most significant regulatory intervention yet in AI platform competition. If enforced, it could fundamentally reshape the mobile AI assistant market — opening the door for companies like Anthropic, OpenAI, and Moonshot to compete on equal footing with Google on Android, the world's most popular mobile OS. Google has already indicated it will appeal, setting up a landmark legal battle over the boundaries of platform power in the AI era.
📖 Related Reads
Want to go deeper on the companies and technologies shaping these stories?
- See how AI infrastructure platforms compare → — Databricks vs Snowflake vs competing data-AI stacks
- Can AI really fly combat missions? — Read our analysis of defense AI →
- The open-weight model boom: How models like Kimi K3 and Qwen3.8 are reshaping the AI landscape — read our analysis
- EU Digital Markets Act explained: What the DMA means for AI competition — read our analysis
That's your AI briefing for July 20, 2026. The big themes this week: China enters the frontier-model race (Qwen3.8 Max), enterprise AI infrastructure keeps printing money (Databricks at $188B), regulators are finally taking AI platform power seriously (EU DMA order), and research keeps reminding us that AI tools change how we think — not always for the better.
📖 Related Reads
- ToolBrain — tool reviews, LLM comparisons, and AI workflow guides
- CodeIntel Log — code quality, debugging, and software engineering benchmarks
- NiteAgent — AI agent development, frameworks, and production patterns
Cross-links automatically generated from None.
Back to all posts
