In partnership with
Welcome back to The AI Field.
Three very different versions of the AI future landed this week. OpenAI's next model, GPT-5.6, is already quietly running inside ChatGPT Pro, and testers say it's fast. Midjourney, the AI image company, announced it's building a 60-second full-body scanner you step into like a spa. And a Chinese open-weights model, GLM-5.2, beat GPT-5.5 on coding for about a sixth of the price.
In this issue:
🤖 GPT-5.6 is already running inside ChatGPT Pro
🧬 Midjourney's 60-second full-body scanner
🇨🇳 China's GLM-5.2 beats GPT-5.5 on coding, at 1/6 the cost
Read time: 5 minutes.
The ones showing up in LLMs convert 3× better than Google
They optimized for LLMs, not just Google.
FAQs. Comparison pages. Transparent pricing. LinkedIn presence. These aren't vanity plays. They're what gets you cited in ChatGPT, Gemini, and Claude when your buyers are researching, your investors are looking, and your future hires are deciding where to work.
Download the free AEO Playbook for Startups from HubSpot and get the exact checklist. Five minutes to read.
The AI Field: OpenAI hasn't announced GPT-5.6 yet, but testers say it's already live inside ChatGPT Pro, and it's quick. With Anthropic's Fable 5 stuck offline under a government ban, OpenAI has a clear lane to retake the lead.
Key details:
OpenAI Chief Scientist Jakub Pachocki has called it a "meaningful improvement" over GPT-5.5. There's no model card and no official date yet.
One tester built a full browser game in about 60 minutes with what they believe is GPT-5.6 Pro. Others report a context window roughly 40% larger than GPT-5.5's 1M tokens.
Polymarket traders put about 83% odds on a launch between June 22 and 28.
Why This Matters: The model race is back to weeks, not quarters. If GPT-5.6 lands the agentic-coding gains testers are seeing, OpenAI reclaims the top spot it lost when Fable 5 launched. The timing, right as its biggest rival is benched, is not an accident.
The AI Field: Midjourney, the company known for AI images, says its next product scans your whole body in about a minute while you float in a pool of warm water. It's the strangest, most ambitious AI announcement of the year, and it's real, sort of.
Key details:
The "Ultrasonic CT" uses sound waves and water, no radiation and no magnets, to build a 3D body map Midjourney claims rivals MRI at nearly 100x the speed.
Reality check: the prototype takes about 20 minutes, not 60 seconds, roughly 12 people have been scanned, and there's no FDA clearance.
A "Midjourney Spa" is planned for San Francisco in late 2027, with a goal of 50,000 scanners doing a billion scans a month by 2031.
Why This Matters: This is AI leaving the screen and reaching for your body. The vision, cheap and constant health data for everyone, is huge. But the gap between cool images and a safe, approved medical product is the entire game. Founder David Holz is selling a 2031 dream, not a 2026 device.
The AI Field: Z.ai (formerly Zhipu AI) released GLM-5.2, an open-weights model that beats GPT-5.5 on long, autonomous coding tasks and costs a fraction of the price. The catch for US labs: anyone can download it for free.
Key details:
On the FrontierSWE coding benchmark it scored 74.4%, ahead of GPT-5.5's 72.6% and nearly level with Claude Opus 4.8 at 75.1%.
It runs at roughly one-sixth the cost of GPT-5.5, with open weights under an MIT license on Hugging Face.
It's free to try at chat.z.ai, and enterprises can self-host it.
Why This Matters: The US export ban pulled Fable 5 offline to slow China down. Instead, a Chinese open model is now co-leading on coding and undercutting everyone on price. When the best value in frontier AI is free and Chinese, the moat isn't the model anymore.
🛠️ TOOL OF THE WEEK
Google AI Studio: the free playground for testing AI models
With a new model dropping almost every week, AI Studio lets you run and compare Google's Gemini models (including the new 3.5 Flash) side by side, tweak settings, and build prompts, all free with a Google login.
Best for: anyone who wants to test models before paying for one.
Price: Free at aistudio.google.com.
Four tools on our radar this week that are worth checking out.
💻 OpenCode - the open-source AI coding agent developers are switching to, works with 75+ models, free.
📝 Granola - takes AI meeting notes straight from your system audio without joining the call as a bot, free plan.
🎨 Krea - an AI creative suite that bundles 60+ image and video models in one place, free daily credits.
⚙️ n8n - build AI automations and agents by connecting your apps, open-source and free to self-host.
🔥 THE AI FIELD'S TAKE
These three stories reveal how AI is advancing in several directions at once.
OpenAI is pushing the limits of raw capability with GPT-5.6. China’s GLM-5.2 is competing aggressively on price, offering a model that reportedly comes close to the industry’s best. Meanwhile, Midjourney is moving beyond the screen altogether, exploring technology that can scan the human body.
For the past three years, the AI race has largely revolved around one question: who has the smartest chatbot? That question is becoming less relevant.
The new frontier is a competition across three areas: capability, affordability, and integration into the physical world.
The most intelligent model still matters. But the bigger shift is that AI is becoming cheaper, more unconventional, and far more personal, all at the same time.
FERC ordered grid operators to fast-track power for AI data centers. The US energy regulator called speeding up grid access a "national priority."
Amazon shelved its finished Sam Altman movie "Artificial." The Luca Guadagnino film was dropped, reportedly to protect Amazon's $50B OpenAI deal.
ChatGPT fell below 50% of AI assistant users for the first time. It's at 46.4%, with Gemini at 27.7% and Claude at 10.3%.
10% of people now use AI chatbots for news every week. That's up from 7%, but only 4% click through to the original source.
Salesforce is buying AI support startup Fin for $3.6B. The former Intercom will power its Agentforce customer-service agents.
📲 TRENDING IN AI
Midjourney's body-scanner reveal blew up on X and Hacker News, with people split between "this is the future" and "I am not bathing in a golden AI pool."
Testers flooded X and Reddit claiming a mystery model in ChatGPT Pro built a full browser game in about an hour, fueling GPT-5.6 hype. Polymarket has it at roughly 83% to ship this week.
GLM-5.2's benchmark wins went viral with one recurring jab: the US banned Fable 5 to slow China, and a free Chinese model promptly took the lead.
💡 PROMPT OF THE WEEK
The highest-leverage prompt there is: make the AI interview you before it answers. It's the one trick that reliably turns generic output into something tailored to you.
Tool: ChatGPT, Claude, or Gemini
Prompt: "You're a world-class expert at [your task, e.g. writing a cover letter, planning a product launch, naming a company]. Before you write anything, ask me clarifying questions one at a time until you have everything you need to do this really well. When you have enough, tell me, then give the final result."
What you'll get: instead of a generic first draft, the AI pulls the important details out of you and delivers something tailored, usually right on the first try.
👀 WHAT TO WATCH
GPT-5.6: Polymarket puts about 83% odds on an official launch between June 22 and 28. Watch for the drop.
Gemini 3.5 Pro: Google promised a June launch. It's still enterprise-preview only, with the month nearly gone.
Midjourney: the first San Francisco spa is targeted for late 2027, with FDA submissions ahead. The hard part starts now.
WRAP-UP
That's all for today's roundup!
🤖 GPT-5.6 is already running inside ChatGPT Pro, with a launch expected any day.
🧬 Midjourney unveiled a 60-second full-body scanner and a spa to put it in.
🇨🇳 China's GLM-5.2 beat GPT-5.5 on coding at a sixth of the cost.
What did you think of today’s edition?
Until next time!
Olle | Founder of The AI Field