Good morning, AI enthusiasts. A 6-person AI startup just did what Google's massive AI division couldn't, beat Gemini 3 Deep Think on abstract reasoning at half the price.
Poetiq's 54% score on ARC-AGI-2 marks the first time any system broke the 50% barrier, and they achieved it 173 days after founding without training their own models. The efficiency play is getting real.
In today's recap:
Poetiq's 6-person team beats Gemini 3 on reasoning
OpenAI shows 8x enterprise growth amid Google pressure
Pope warns AI risks concentrating power
5 new AI tools and more
POETIQ
6-person startup tops Google on reasoning
Source: Poetiq
Recaply: Poetiq just surpassed Gemini 3 Deep Think on ARC-AGI-2 with 54% accuracy versus 45%, first system to break 50% on abstract reasoning.
Key details:
Poetiq's meta-system learns to solve tasks by creating systems utilizing any frontier model, improving with every task through understanding solutions.
The company scored 54% at $30.57 versus Gemini 3's 45% at $77.16, establishing state-of-the-art verified by ARC Prize.
Six researchers with 53 combined DeepMind years built the solution without training their own models, achieving results within hours of Gemini 3 release.
Poetiq topped the official leaderboard with a publicly released open-source solver evaluated on the Semi-Private Test Set, and is seeking partners for real problems.
Why it matters: The $1.4 trillion AI infrastructure race assumes intelligence requires massive scale, but Poetiq beat Google's model without competing infrastructure. Companies race to secure chips and capacity, and this shows optimization may deliver more value than raw improvements. Whether meta-systems scale beyond benchmarks determines if efficiency competes with scale industrywide.
OPENAI
OpenAI touts growth amid Google threat
Source: OpenAI
Recaply: OpenAI just released “the state of enterprise AI” reports showing 8x message growth days after CEO Sam Altman sent internal code red memo about Google's competitive threat.
Key details:
Report shows ChatGPT Enterprise serves 7 million workplace seats with seats up 9x year-over-year, while 36% of US businesses use ChatGPT Enterprise versus 14.3% for Anthropic.
Workers report saving 40 to 60 minutes daily, Custom GPTs usage jumping 19x year-to-date and processing 20% of Enterprise messages.
Technology companies lead API usage at 5x growth for customer applications, non-technology firms grew 5x, with customer service and content now 20% of activity.
Timing follows Altman's code red warning about Google as majority of OpenAI revenue comes from consumer subscriptions threatened by Gemini despite $1.4 trillion infrastructure commitment.
Why it matters: After internal alarms about Google and pressure from Anthropic's B2B success, OpenAI needed to demonstrate enterprise momentum is real. Data shows genuine adoption across industries, but releasing growth report during competitive crisis feels defensive. OpenAI views this as proving leadership, but widening gap between frontier and median users suggests most haven't extracted full value, creating opening for competitors.
VATICAN
Pope warns AI concentrates power
Source: Vatican Media
Recaply: Pope Leo XIV warned artificial intelligence risks accumulating wealth and power in few hands, urging humanity ensure technology serves common good during Vatican address.
Key details:
At a Vatican conference titled "Artificial Intelligence and Care of Our Common Home," the pontiff addressed the Centesimus Annus Pro Pontifice Foundation and Catholic Research Universities alliance, calling for technology that serves humanity rather than exploits it.
Leo XIV emphasized humans must be active co-workers in creation versus passive consumers of AI content, highlighting concerns about critical thinking, discernment, and relationships.
Remarks coincided with December 2 UN Development Programme report warning unmanaged AI risks sparking new divergence era by widening inequality between wealthy and developing nations.
Pope stressed concern for children whose freedom, spirituality, and development face vulnerability, calling for widespread participation in AI governance.
Why it matters: Pope Leo XIV calls for equitable AI alongside activists and academics, but absence of tech leaders signals a problem. It's hard to ensure technology serves common good when companies building systems aren't joining conversations. Time will tell if religious frameworks influence development or if market forces prove stronger than calls for restraint.
NEWS
Beyond the big stories in AI today
The New York Times and Chicago Tribune sued Perplexity AI for copying and distributing exclusive content.
Trump approved Nvidia H200 chip sales to China with 25% government cut, potentially worth $2B to $5B quarterly in additional revenue.
Elon Musk announced xAI will launch Grok 4.20 AI model within 3 to 4 weeks, marking one of fastest upgrade cycles.
OpenAI partnered with Instacart to enable grocery shopping and checkout directly in ChatGPT through new Agentic Commerce Protocol integration.
Anthropic launched Claude Code integration in Slack, automatically routing coding tasks from Slack messages to web-based Claude Code sessions in beta.
Z AI released GLM-4.6V models with 106B flagship and 9B Flash versions, featuring first native Function Calling in vision model family.
Google announced AI-powered smart glasses with Gemini launching in 2026, partnering with Samsung, Warby Parker, and Gentle Monster on hardware.
IBM acquired Confluent for $11B to create smart data platform for enterprise AI, expected accretive to adjusted EBITDA within first year.
TOOLS
Trending AI Tools
🎤 ACE Studio 2.0 - The all-in-all AI music canvas
📊 Paper2Slides - Transform research papers & technical reports into professional presentations
🎨 Mixboard - AI-powered moodboard builder from Google for visual brainstorming
📝 Dapple - Submission management software for creative industries
🤖 GLM-4.6V - Multimodal AI model with native tool use
