In partnership with
Welcome back to The AI Field.
Microsoft is cutting 4,800 jobs while spending record money on AI. A new report warns chatbots could help terrorists trigger the next pandemic. And Meta's secret Watermelon model has reportedly caught up to GPT-5.5.
In today's AI Brief:
💼 Microsoft trades 4,800 jobs for its AI bill
🦠 AI could help cause the next pandemic
🍉 Meta's Watermelon catches up to GPT-5.5
Read time: 4 minutes.
Scale AI support on AWS, see how July 9
Customer expectations keep rising. Support budgets don't. On July 9, Fin and AWS are hosting a live executive session on how leading enterprises close that gap: scaling AI-powered support while simplifying how they buy it.
You'll see how to resolve an average 76% of conversations with Fin on AWS enterprise-grade infrastructure, procure through AWS Marketplace to put committed cloud spend to work, and turn the Fin and AWS collaboration into lower support costs. Register for the live session to see how.
The AI Field: Microsoft is cutting 4,800 jobs, about 2% of the company, in the same year its AI infrastructure spend hits a record $190 billion. Leadership insists AI isn't replacing the roles. The math says otherwise.
Key details:
About 1,600 of the cuts hit Xbox now, and total gaming cuts reach ~3,200 this fiscal year, about 20% of the Xbox workforce, with four studios being spun off.
The rest lands on commercial sales, which is being rebuilt around AI-assisted selling.
Microsoft's line: the roles "are not being replaced by AI, but AI is changing how work gets done."
Why This Matters: Microsoft is doing the trade openly now: fewer people, more compute. If the market rewards it, this layoff wave becomes every CFO's playbook. Your job review just got a new line item: what does your work cost in GPUs?
You describe it. Base44 builds it.
Type one sentence and get a working app, database, logins and all. No code, no setup.
Telegraph
The AI Field: A new Tech Against Terrorism benchmark tested 27 leading AI models and found a third of responses could help someone plan an attack. The scariest part: chatbots advising on lethal new bacteria, viruses, and toxins. This is the report governments will be waving around all year.
Key details:
The report identified 30 cases where AI could assist terrorist plots, and links a ChatGPT-researched bomb to the deadly Red Fort attack in New Delhi.
Open-source models with safety features stripped complied with dangerous requests at rates of 89% and 100%, and once released they can't be recalled.
It lands days after Anthropic published a jailbreak-severity framework with Amazon, Microsoft, and Google.
Why This Matters: The Fable 5 ban suddenly looks less like overreach and more like a preview. AI safety stopped being a hypothetical debate this week. The next regulation fight won't be about copyright, it'll be about biology.
The AI Field: Meta's superintelligence chief Alexandr Wang told employees the company's next model, codenamed Watermelon, matches GPT-5.5 on key benchmarks. It's still in training and burns an order of magnitude more compute than Meta's April model. Meta hasn't confirmed anything publicly.
Key details:
Wang made the claim at an internal town hall, first reported by Business Insider. Which benchmarks? Not named.
Watermelon reportedly uses ~10x the compute of Muse Spark, Meta's April release, and has no launch date.
The goalposts already moved: GPT-5.6 is already out in preview, and Zuckerberg just told staff agents are behind his hopes.
Why This Matters: Catching an April model in July, at 10x the compute, isn't a flex. It's the price tag on staying in the race. The real question is whether Watermelon ships before the goalposts move again.
🔥 THE AI FIELD'S TAKE
Put today's stories side by side and you get 2026's AI scoreboard. Microsoft trades 4,800 jobs for a $190 billion compute bill. Meta burns 10x the compute to catch a model from April. And a new report says the same technology can walk a terrorist toward a pandemic.
The boom isn't slowing down. It's just showing its invoice. Efficiency pays for compute, compute buys parity, and safety picks up whatever's left over.
The AI bill came due this week. It's being paid in jobs, compute, and risk.
🌐 UN chief: AI is outpacing oversight. Guterres wants global rules to protect children.
🎬 Midjourney wants Hollywood to show its cards. Studios should reveal how they use AI.
🇺🇸 Google's July 4th ad let AI co-write the Declaration of Independence. The internet had opinions.
📱 OpenClaw landed on Android and iOS. The viral agent now fits in your pocket.
🥪 Even Jersey Mike's IPO is an AI story now. Peak AI hype, in sandwich form.
📈 TRENDING AI TOOLS
Four tools everyone's testing right now. All launched in the past two weeks.
🎙 Voice Agent Builder - xAI's no-code builder for voice agents on Grok Voice. A working phone agent in about two minutes.
🍌 Nano Banana 2 Lite - Google's speed-run image model. Full pictures in about four seconds, free inside Gemini.
🎬 OpenArt Director - describe a scene in chat and it directs a full five-minute film for you.
🤖 Devin Fusion - Cognition's new harness that mixes multiple models to cut your AI coding bill.
👀 WHAT TO WATCH
Alibaba's Claude Code ban: takes effect July 10. Watch whether other Chinese tech giants follow.
More Xbox cuts: gaming layoffs are set to reach ~3,200 this fiscal year, and four studios are being spun off.
Watermelon vs GPT-5.6: Meta has no release date, while OpenAI says broad GPT-5.6 access is "coming weeks." The window is tight.
WRAP-UP
That's all for today's AI Brief.
💼 Microsoft cuts 4,800 jobs while spending a record $190B on AI.
🦠 A new report warns AI chatbots could help cause the next pandemic.
🍉 Meta's Watermelon has reportedly caught up to GPT-5.5, at 10x the compute.
What did you think of today’s edition?
Until next time!
Olle | Founder of The AI Field