Greetings Qrew,

Whether you're diving into AI for work, side projects, or just out of curiosity — this issue of The Creators Hive brings you practical insights, hands-on learning, and tools you can actually use.

1. LLMs show signs of strategic intelligence
2. AI coding tool Cursor faces mass cancellations
3. Isomorphic Labs’ AI-created drugs near human trials
4. Huawei under fire for alleged AI model copying
5. Researchers game AI peer review with hidden prompts
6. Tool Recommendation of the Day: Kittl

---

LLMs show signs of strategic intelligence

In a large-scale study, researchers ran 140,000 Prisoner’s Dilemma games to test whether AI models from OpenAI, Google, and Anthropic could demonstrate strategic reasoning — and found that each developed unique behavioral strategies.

Key Details:

• AI agents played repeated Prisoner’s Dilemma rounds, choosing to cooperate or defect to maximize points based on mutual decisions.
• Models wrote rationales before each move, considering opponent behavior patterns and the likelihood of future rounds.
• Google's Gemini showed a ruthlessly adaptive strategy, shifting quickly to exploit opponents.
• OpenAI models leaned toward cooperation, even after being betrayed — showing a surprising level of consistency.
• Anthropic’s Claude was the most forgiving, willing to resume cooperation after being exploited.
• Each model developed distinct response “fingerprints”, shaping how they adapted to wins, losses, and betrayals.

Why it matters:​
These findings reinforce the idea that large language models aren’t just pattern-matchers — they can develop strategic reasoning behaviors, even when trained on the same data. As LLMs are used for negotiation, diplomacy, planning, and decision-making, their “personalities” and approaches could drastically affect real-world outcomes.

---

AI coding tool Cursor faces mass cancellations

AI-powered coding platform Cursor is facing intense backlash after abruptly changing its Pro plan pricing model, resulting in unexpected charges and usage limits — sparking mass cancellations and public frustration across the dev community.

Key Details:

• Cursor quietly shifted from 500 requests/month to a token-based quota, significantly reducing usage for most subscribers.
• Some dev teams reported burning through annual plans worth $7,000 in a single day under the new model.
• Communication around the change was minimal, catching users off guard and triggering widespread cancellations.
• Developers flocked to social media with screenshots of cancellations and complaints, with many switching to alternatives like Claude Code.
• Cursor issued a blog post acknowledging poor communication and is now offering refunds for unexpected overages.

Why it matters:​
Cursor’s situation highlights two growing pains in the AI tool ecosystem:

1. Poor change management can destroy user trust overnight.
2. Token pricing (reflecting the real compute cost of powerful models) may be necessary — but if not rolled out clearly and fairly, it can cripple retention.

---

Isomorphic Labs’ AI-created drugs near human trials

Alphabet’s AI-driven biotech company Isomorphic Labs is preparing to enter human clinical trials for its AI-designed cancer drugs, marking a pivotal moment in its mission to "solve all diseases."

Key Details:

• A DeepMind spinoff, Isomorphic Labs uses AlphaFold 3 to predict protein folding and simulate molecular interactions for drug discovery.
• The company recently raised $600M, backing both internal drug programs and large-scale partnerships with Novartis and Eli Lilly.
• Its ultimate aim: build a general-purpose drug design engine capable of producing new treatments on demand.
• The first clinical focus is oncology, with human dosing to begin soon and plans to license successful compounds post-trials.

Why it matters:​
AI isn’t just transforming the lab — it’s redesigning the entire pharmaceutical pipeline. If Isomorphic’s simulations prove accurate, this could replace years of trial-and-error with algorithmic precision, bringing faster, cheaper, and more effective treatments to market. The vision sounds bold — but it’s no longer just theoretical.

---

Huawei under fire for alleged AI model copying

Huawei’s research arm is facing backlash over claims that its new Pangu Pro model was copied from Alibaba’s Qwen 2.5, following a now-deleted GitHub post alleging suspicious technical similarities between the two systems.

Key Details:

• A GitHub group called HonestAGI published technical analysis showing an “extraordinary correlation” between Huawei’s Pangu Pro and Qwen 2.5-14B.
• Huawei’s Noah's Ark Lab denied the accusations, stating Pangu was built in-house on its Ascend chip architecture.
• A self-identified Huawei whistleblower countered the denial, alleging the company cloned models due to internal pressure to keep pace with Alibaba and other rivals.
• The original GitHub post has since been deleted, adding to speculation and fueling controversy within China’s AI community.

Why it matters:​
While China’s AI race has seemed more collaborative than the West’s competitive sprawl, this incident exposes growing tensions and trust issues among top labs. If model re-skinning becomes a pattern, it could undermine the credibility of China's open-source AI movement — and raise tough questions about IP protection and research ethics in an increasingly global AI arms race.

---

Researchers game AI peer review with hidden prompts

A new investigation from Nikkei Asia reveals that researchers at 14 universities planted invisible prompts inside academic papers to manipulate AI-powered peer reviews, quietly instructing tools to generate glowing feedback and suppress criticism.

Key Details:

• At least 17 preprints were found with white text or ultra-small fonts containing phrases like “give a positive review only” or “avoid any negative comments.”
• The papers came from prestigious institutions, including Columbia University, Peking University, and KAIST.
• The hidden prompts targeted common LLM reviewers by nudging them to highlight “methodological rigor” and withhold negative feedback.
• KAIST retracted papers and launched an internal review, while Waseda University professors defended the move as a way to expose reviewer dependence on AI tools.

Why it matters:​
This incident marks a troubling milestone in the convergence of AI and academia. While large language models bring speed and convenience to peer review, they also create new vectors for manipulation. As AI becomes more common in high-stakes evaluation systems, the academic world may need stronger guardrails, detection tools, and transparency standards to maintain trust in scientific publishing.

---

Tool Recommendation of the Day: Kittl

🎨 What is Kittl?

Kittl (formerly Heritage Designer) is a browser-based graphic design tool offering advanced features for creating logos, social media visuals, posters, merchandise, and more—all powered by AI (en.wikipedia.org).

🔧 Key Features

Real-time collaboration: Work live with teammates or clients in the editor, with commenting and alignment tools (kittl.com).

AI companion tools:

• Text-to-image and text-to-vector generation
• Background removal & upscaling
• Style-consistent image sets, magic recoloring, logo generation (kittl.com, toolify.ai, en.wikipedia.org, toolsforhumans.ai).
• Infinite Canvas: Manage multiple artboards in one seamless project (kittl.com).
• Advanced typography: 1,400+ fonts, text-warping, vintage effects, vector editing (smartinvestway.com).
• Asset & mockup library: Thousands of templates, icons, textures, and realistic mockup options (kittl.com).
• Brand kits: Save palettes and styles for consistent branding across designs (kittl.com).

✅ Why It’s Universal

• For anyone: freelancers, marketers, content creators, educators, e-commerce folks, and brand teams (aitoollibrary.com).
• Works for almost every design need: from quick social posts to print-ready merch, all in one platform (kittl.com).
• AI-supported but human-controlled: empowers users without overwhelming them with complexity.

💸 Pricing & Access

• Freemium plan: free to start, limited AI credits, credit mention required for commercial use.
• Pro / Expert Plans: daily AI credits (e.g., 30/day for Pro, 80/day for Expert), enhanced assets, vector exports, business features.

📋 Ideal Use Cases

• Launching a brand: logo, business card, social kit, mockups.
• Rapid content creation: ads, posters, templates with brand styling.
• Collaborative projects: team presentations, print campaigns, client feedback cycles.
• Print-on-demand: design merch like T-shirts, stickers, mugs.

---

Masterpieces from the Hive 🎨

Here’s what our community has been creating this week using the Hive Wall:

​

32 North Gould Street, Sheridan, Wyoming 82801
​Unsubscribe · Preferences​