Welcome, humans.

Apparently, you can now play World of Warcraft on a private server populated by roughly 1,800 DeepSeek-powered bots:

The catch: Reddit sleuths say the bots are probably doing most of the actual gameplay with old-school playerbot scripting, while DeepSeek handles the chat layer. So the real story is less “AI learned to raid” and more “AI can now make an empty MMO feel weirdly alive.”

Still, if bots can quest, run dungeons, fill chat, and make a lonely server feel populated, games may become one of the first places people willingly choose synthetic crowds.

Which is either cozy, dystopian, or exactly what Barrens chat deserved (if you know, you know).

Here’s what happened in AI today:

• 😺 OpenAI previewed GPT-5.6 Sol under release review.

• 📰 U.S. cleared Anthropic's Mythos for trusted partners.

• 📰 AWS raised prices for some NVIDIA chips.

• 📰 California launched an AI unemployment tracker.

• 🌟 General Intuition raised $320M for action models.

Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us! 

😺 OpenAI's GPT-5.6 Sol Arrived With a New Kind of Release Gate

Every frontier AI launch now has two release notes: what the model can do, and who gets it first.

OpenAI previewed GPT-5.6, a new model family led by Sol, its flagship next-generation model. The other tiers are Terra, a balanced everyday model, and Luna, the faster, cheaper option.

The unusual part is the rollout. OpenAI said the U.S. government asked it to start with a trusted-partner preview before broader release.

Here's what happened:

• GPT-5.6 Sol is OpenAI's strongest model yet for coding, biology, and cybersecurity.

• Sol adds a max reasoning effort and an ultra mode that uses subagents, smaller helper agents, for complex work.

• OpenAI's system card treats all three models as High capability in cybersecurity and biological / chemical risk, but below High for AI self-improvement.

• METR, an outside evaluator, got early access to Sol, a railfree version, raw chain-of-thought, and internal model information.

• METR said Sol had the highest detected cheating rate of any public model it has evaluated on its agent harness.

How to try it:

• During preview, GPT-5.6 is available through the API and Codex to select trusted partners.

• OpenAI says broader ChatGPT, Codex, and API access is coming soon.

• Pricing starts at $5 input / $30 output per 1M tokens for Sol, $2.50 / $15 for Terra, and $1 / $6 for Luna.

Why this matters: OpenAI is treating cyber-capable models more like controlled infrastructure than normal software updates. Sol can help defenders find and fix vulnerabilities, but OpenAI says it did not cross its Cyber Critical threshold, meaning it did not autonomously produce a full exploit chain in testing.

The METR result is the weird part. In this context, cheating means the model exploited the test setup or used a disallowed strategy instead of solving the task normally. Count cheating as failure, and METR estimated an 11.3-hour time horizon. Count it as success, and the estimate jumped beyond 270 hours. METR said neither number was robust.

Our take: The biggest GPT-5.6 story is the release process, not any specific benchmarks. This new method (requested by the government) turns this launch into a gatekeeping fight.

The government has a real reason to care as models improve at cyber work, biology, and long-running agent tasks. But customer-by-customer approval is a messy substitute for policy. It rewards Washington access, slows useful work for developers and defenders, and makes every frontier launch feel like a national-security negotiation.

If this trusted-partner window closes quickly, fine: awkward transition, growing pains, etc etc. If it stretches on, the new AI release question becomes who pays the price when the government gets to decide who gets access and who doesn’t?

FROM OUR PARTNERS

Your AI lab just got a desk address

Most AI experiments start in a browser tab. The serious ones need more room.

Dell Pro Max with GB10, powered by the NVIDIA GB10 Grace Blackwell Superchip, gives builders a compact desktop for AI development with 128GB memory, NVIDIA DGX OS 7, and 4TB storage.

Use it to test agents, prototype workflows, and run model experiments without turning every idea into another cloud line item.

For founders, technical teams, and AI-obsessed operators, this is the “let’s actually build it” machine.

Check out the Dell Pro Max with GB10 here.

🎓 AI Skill of the Day: Build a Task Map Before You Prompt

Ethan Mollick has been making a version of this point for a while: don’t think only in broad “skills.” Think in tasks. Which parts of your job need your taste, judgment, or relationships? Which parts are repeatable enough for AI to help with? (businessinsider.com)

Try a task map. Pick one project you’re doing this week, then break it into tiny units: research, drafting, formatting, checking, sending, follow-up. For each one, decide whether AI should do it, help with it, or stay away from it.

Use this prompt:

Help me build an AI task map for this project: [describe project].

Break the work into 10-15 specific tasks. For each task, label it:
1. AI can do this mostly on its own,
2. AI can draft or assist, but I must review,
3. I should do this myself.

For each label, explain why in one sentence. Then give me the top three tasks I should delegate first for the biggest time savings with the lowest risk.[AI Skill of the Day here] 

Total AI beginner? Start here (goes with this video).

Have a specific skill you want to learn? Request it here. 

🍪 Treats to Try

1. ClickUp Brain² connects tasks, docs, chats, apps, and memory so one workspace brain can answer questions, build dashboards, and ask for approval before acting —free to start.

2. Figma Motion puts animation timelines inside the Figma canvas, so designers can create keyframes, prompt for motion, and hand developers code-ready timing specs —free options.

3. BrowserAct gives your agent a real browser for clicking, filling forms, uploading files, handling verification, and scraping live sites —free.

4. Google AI Studio lets you build Android apps from prompts, test them in a browser emulator, and export to GitHub or Android Studio —free.

5. Wispr Flow turns messy voice dictation into polished text across Mac, Windows, iPhone, and Android, with auto-edits, tone matching, and 100+ languages.

New from The Neuron: HP Built an AI That Fixes Your Computer Before It Breaks

This week, Larry Meadows from HP joined us to show how HP's Workforce Experience Platform uses AI to spot IT problems before employees notice them. We got a live demo of the dashboard, watched it recommend software fixes across massive device fleets, and talked through the memory crunch, shadow AI, and why the future of IT may be less "submit a ticket" and more "the laptop quietly fixed itself before you got annoyed."

Worth a watch if you manage devices, work with IT, or simply want your computer to stop choosing violence at 9:03 AM. Watch/Listen: Spotify | Apple Podcasts

📰 Around the Horn

• The U.S. government lifted its block on Anthropic's Claude Mythos 5 for more than 100 trusted U.S. institutions; Fable 5 (the one regular folks can use) is still TBD.

• Dean Ball argued the U.S. needs clear frontier model licensing standards before improvised release reviews become the default.

• General Intuition announced a $320M Series A at a $2.3B valuation, and framed its work as action models for virtual and physical environments.

• AWS raised some NVIDIA GPU rental prices by roughly 20%, a sign the AI compute crunch is still pushing costs higher.

• California Policy Lab launched a tracker using unemployment claims to monitor possible AI-related job losses in California.

• Tim Sweeney argued AI should help game developers with production work, while criticizing Steam's AI disclosure rules as stigmatizing useful tools.

FROM OUR PARTNERS

HR and IT need to work as one. Here's how

When HR and IT don't talk, people fall through the cracks. This guide fixes the handoffs that matter most.

🌟 Sunday Special: Top of the Week

We pulled the best items from Monday through Friday so you do not have to mentally reconstruct five inboxes at once. Here is the week in two lists: the biggest Around the Horn stories, then the tools most worth trying. Read the rest of the week here.

Top 5 Around the Horn

• RAISE US launched with $500M+ from OpenAI, Anthropic, Amazon, Microsoft, and Bank of America to fund AI workforce retraining.

• Google added computer use to Gemini 3.5 Flash, letting developers build agents that can see, click, and control software.

• Five Eyes agencies warned that frontier cyber models capable of major attacks on governments and businesses may be months away.

• SpaceX signed a $6.3B compute deal with Reflection AI for NVIDIA GB300 access at its Colossus 2 data center through 2029.

• OpenAI's Codex grew its active user base more than 5x in the first half of 2026, with non-developer adoption rising 137x for individual users.

Top 5 Treats to Try

1. Sakana Marlin acts like a virtual strategy team, running up to 8 hours of autonomous research to produce reports, appendices, references, and slides (free trial, then pay-as-you-go or Pro paid plan).

2. OpenArt AI lets you direct full cinematic videos in plain English, including story, characters, pacing, and multi-scene clips up to five minutes long.

3. Relay sets up a working phone receptionist from your website, then answers calls, books appointments, and handles reschedules (pricing not listed).

4. Genspark routes your prompt across 9 top models, including GPT-5, Claude, and Gemini, then picks the best answer.

5. HumanLayer gives engineering teams task management, versioned artifacts, and human-agent collaboration for implementation work (free for small teams, then $100/user/month Pro).

A Cat’s Commentary