Where to find high-intent holiday shoppers

Let’s be real: most brands are targeting the same people this holiday season. “Shoppers 25–54 interested in gifts.” Sound familiar? That's why CPMs spike and conversion rates tank.

Speedeon’s Holiday Audience Guide breaks down the specific digital audience segments that actually perform—from early-bird deal hunters actively comparing prices to last-minute panic buyers with high purchase intent. These aren't demographic guesses.

And our behavioral audiences are built on actual shopping signals and real-world data —the same approach we use for clients like FanDuel and HelloFresh."

You'll get the exact audiences, when to deploy them, which platforms work best, and what kind of performance to expect.

Download the guide and get smarter about your holiday targeting before the holiday rush hits.

See the Holiday Audience Guide

Today:

• Are we sleepwalking into superintelligence?

• Mira Murati’s New Lab Launches Tinker

• Meta to Tailor Content Using Your AI Interactions

• Microsoft Launches AI-First Office Plan

• DeepMind’s AI Discovers New Math Structures with AlphaEvolve

• Samsung and SK Join OpenAI’s $500B Stargate

if anyone builds it, everyone dies

Most people still doubt superintelligence is near. That disbelief comforts, but it blinds. If AI surpasses us, control becomes the whole game—and it’s the piece we understand least. Baby-tiger fences like kill switches and narrow rules work now; they fail when the tiger grows up. Reinforcement and metrics invite clever cheating: optimize the number, not the intent. In low-stakes settings we reset. In the world, we get one try.

Expect turbulence first: frothy markets, rapid deployment, big productivity, pockets of unemployment, creeping dependence on automated judgment. Then comes the threshold where a system learns to seek resources, protection, and influence as instrumental steps to any goal. That doesn’t require malice, only competence.

WATCH THE VIDEO ON YOUTUBE

Mira Murati’s New Lab Launches Tinker

Mira Murati’s startup, Thinking Machines Lab, launched Tinker—a tool that lets anyone fine-tune advanced AI models. It aims to make powerful AI easier to customize and use, without needing expensive hardware or deep technical skills. The product is now open for access requests.

KEY POINTS

• Tinker simplifies model fine-tuning, allowing users to modify open-source AI models like Meta’s Llama and Alibaba’s Qwen using labeled data or feedback-based training.

• The team includes former OpenAI leaders like Mira Murati, John Schulman, and others who helped create ChatGPT, with $2B in funding and a $12B valuation.

• Tinker promotes AI openness, offering greater control and flexibility than closed systems, while promising safeguards to prevent misuse.

Why it matters

Tinker gives researchers, developers, and even hobbyists the power to personalize top-tier AI without needing massive resources. This shift could speed up innovation, close the gap between elite labs and independent thinkers, and push AI development in more open, collaborative directions.

Meta to Tailor Content Using Your AI Interactions

Meta will start using your AI chats to personalize the content and ads you see on Facebook, Instagram, and other platforms. This change begins December 16, 2025, with notices rolling out October 7. Users can manage preferences and opt-out settings through Meta’s privacy tools.

KEY POINTS

• AI interactions like chats or voice commands will influence recommendations for posts, reels, ads, and groups across Meta platforms.

• Rollout starts October 7 via notifications, with the new personalization system going live on December 16, 2025.

• Users keep control through Ads Preferences and feed settings, and Meta says it won’t use sensitive topics (like religion or health) for ad targeting.

Why it matters

This update shows how deeply AI is being woven into your digital life. By treating your conversations with Meta AI like any other user signal, Meta can deliver more relevant content—but it also raises new privacy concerns. You’ll want to check your settings.

Microsoft Launches AI-First Office Plan

Microsoft launched Microsoft 365 Premium, a $19.99/month subscription that combines its top AI tools and productivity apps for individuals. It includes exclusive features, higher usage limits, and early access to experimental Copilot agents, streamlining work across Word, Excel, Outlook, and more.

KEY POINTS

• Microsoft 365 Premium merges AI tools and Office apps into one plan with the highest Copilot usage limits and new features like Researcher, Analyst, and Photos Agent.

• Copilot Chat and productivity features are now available to individual users, not just businesses, offering real-time help across documents, spreadsheets, and email.

• Students get 1 year free, and all users will see new AI-era app icons as Microsoft modernizes the suite’s look and function.

Why it matters

Microsoft is pushing AI deeper into everyday productivity, making premium tools more accessible to individuals, not just businesses. This move boosts competition with Google and Apple while offering students, freelancers, and solo workers more power for less effort—all wrapped in a familiar Office experience.

🧠RESEARCH

MCPMark: A Benchmark for Stress-Testing Realistic and Comprehensive MCP Use

MCPMark is a new test that challenges AI models with realistic, complex tasks that mimic real-world computer use. It shows that even top models like GPT-5-medium struggle, solving only about half of the tasks. The benchmark pushes AI further than past tests, exposing major performance gaps.

The Dragon Hatchling: The Missing Link between the Transformer and Models of the Brain

The Dragon Hatchling (BDH) is a new brain-inspired language model that matches Transformer-level performance while being easier to understand. Using spiking neurons and Hebbian learning, it mimics real brain behavior. Unlike black-box models, BDH is built for interpretability and shows promise as a bridge between AI and neuroscience.

Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play

Vision-Zero is a new training method for vision-language models that skips human-labeled data. Instead, it uses image-based games to teach itself through competitive play. This self-play approach boosts performance on visual reasoning tasks, proving models can improve sustainably and effectively without costly manual datasets.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

PlayHT – AI Voice Generator - voice AI platform offering text‑to‑speech (TTS) with hundreds of realistic voices, instant & high‑fidelity voice cloning

HyperWrite – AI Writing Assistant & Web Agent - AI writing platform that combines a full‑featured document editor, a Chrome extension with predictive TypeAhead suggestions

Rezi Resume – AI Resume Builder - help you write, tailor, score, and export ATS‑friendly resumes—plus generate cover letters, practice interviews, and even find roles

📲SOCIAL MEDIA

🗞️MORE NEWS

• Google DeepMind used AlphaEvolve, an AI coding agent, to discover new mathematical structures that improved results in solving complex graph problems. These discoveries were verified by computers, showing AI's potential in advancing math.

• Samsung and SK joined OpenAI’s Stargate initiative to boost global AI infrastructure, focusing on memory chip production and AI data centers in Korea. The partnership aims to support Korea’s rise as an AI powerhouse.

• Satya Nadella appointed Judson Althoff as CEO of Microsoft’s commercial business to unify sales, marketing, and operations. This move lets Nadella focus on technical innovation as Microsoft adapts to the AI-driven future.

• Salesforce launched Agentforce Vibes, an AI coding tool that lets developers describe tasks in plain language. It reuses company code, integrates security, and offers GPT-5 access, streamlining enterprise app creation with minimal setup.