Large language models might actually be more than just next word predictors. Anthropic has been putting out incredible papers lately that show AI large language models in particular exhibit very human-like behavior at almost every level. Here's the new paper emergent introspective awareness in large language models.So what did anthropic actually test. There were four main experiment types. First injected thoughts.What they did was use two different prompts, one with all caps and one without all caps. And th ...

AI Agents & Business Transformation - AI agents are poised to transform the workplace by acting on suggestions and driving impact in areas like data entry, invoice creation, and client relationship management [16] - Companies should focus on reimagining their business processes to take full advantage of AI agents, rather than simply automating existing human tasks [43] - Early adopters of AI agents can build compounding intelligence advantages, creating learning gaps over competitors and making their competitive moats increasingly difficult to replicate [48][49] Human-AI Collaboration - To stay relevant, individuals and teams need to develop uniquely human abilities ("humics") such as genuine creativity, critical thinking, and social authenticity to maximize value creation with AI [61][53][54][55][56][57] - Managers need to become orchestrators, designing human-agent workflows and focusing on goal setting and boundary definition, rather than task assignment [66] - Building trust between team members and AI agents is crucial, and can be achieved through clarity, transparency, and a gradual collaboration approach [67][68][69] Implementation Strategies - Successful AI transformations require a clear vision from top leadership and investment in people and talents, including creating a center of excellence or AI talent group [85][86] - Companies should start small with pilot projects to demonstrate the capability of AI agents and generate momentum within the organization [88] - Organizations should allocate approximately 20% of employees' working time to stay informed about AI advancements, experiment with new technologies, and develop AI literacy [80] Metrics & Change Management - Organizations need a comprehensive review of metrics for both AI agents and humans, incentivizing humans to experiment and build uniquely human skills [72] - Companies should combine qualitative and quantitative metrics, focusing not only on cost and efficiency but also on improvements in customer and employee experience [74] - Businesses must cultivate change readiness, enabling them to filter information effectively and identify AI technologies that are worth testing and using [77][78][79]

Performance Overview - The Akre Focus Fund's Institutional share class reported a third quarter 2025 performance of -3.65%, significantly underperforming the S&P 500 Total Return, which was at 8.12% [3][4] - For the trailing 12-month period ending September 30, 2025, the Fund's Institutional share class achieved a return of 3.71%, compared to 17.60% for the S&P 500 Total Return [3] Key Holdings Impact - The primary contributor to the Fund's poor performance was a 26.06% decline in the share price of Constellation Software, which accounted for a 3.56% detriment to the Fund's performance in the quarter [5][6] - Constellation Software has been held by the Fund for over 11 years, with a compound annual rate of return of 24.97% since its initial purchase [7] Long-term Perspective on Constellation Software - Despite the recent drawdown, the Fund maintains confidence in Constellation Software's ability to navigate technological changes and protect its market position [8][9] - Constellation has demonstrated a compound annual total revenue growth rate of approximately 20% over the past decade, indicating strong underlying business performance [8] Leadership Transition - Mark Leonard, the founder of Constellation Software, announced his immediate step down from the President role due to health issues, raising concerns about leadership continuity [12][10] - The Fund expresses optimism regarding Mark Miller, the new President, who has been with Constellation since its first acquisition and is expected to uphold the company's foundational principles [14] Sector and Holdings Composition - As of September 30, 2025, the top five holdings in the Fund included Mastercard (12.4%), Brookfield Corp (10.6%), and Constellation Software (10.1%) [28] - The sector weightings revealed a significant concentration in Financials (52.1%) and Information Technology (20.7%) [28]

Core Insights - OpenAI is exploring partnerships with Broadcom to leverage custom silicon for AI applications, similar to Google's TPU model, which has proven successful in reducing costs significantly [1][2][6] - Broadcom's AI chip business is projected to reach nearly $20 billion, with a substantial portion of revenue coming from Google, indicating a strong market position [1][3] - The cost of chips constitutes 60-70% of data center expenses, making it crucial for companies to source chips at the lowest possible cost to enhance profitability [6][7] Group 1: Cost Efficiency and Custom Silicon - OpenAI aims to reduce costs by 30-40% per gigawatt through the use of Broadcom's chips, which are expected to be cheaper than competitors like Nvidia and AMD [2][3][7] - The integration of custom silicon is essential for optimizing performance per watt, which is a key consideration for AI data centers [9][10] - Google has demonstrated that custom silicon can provide superior performance, as seen in their successful deployment of YouTube videos [9][10] Group 2: Competitive Landscape - Other companies like Amazon and Microsoft are also pursuing custom silicon solutions but have not achieved the same level of success as Google with Broadcom [5][6] - The competitive advantage of Broadcom lies in its ability to offer both merchant and custom silicon tailored to specific needs, which is critical for companies looking to optimize their AI infrastructure [4][7] - The focus on minimizing latency and maximizing performance per watt is driving the demand for custom silicon in the AI sector [9][10]

AI Investment & Infrastructure - OpenAI's AI infrastructure buildout is estimated at $850 billion, potentially requiring the equivalent of 17 nuclear reactors [1] - OpenAI's projects from the past 2 days amount to nearly half of the $2 trillion global AI infrastructure surge forecast by HSBC [2] - Alibaba is increasing its AI model and infrastructure spending, adding to the existing $53 billion commitment [6] - Alibaba is opening new data centers in Brazil, France, and the Netherlands to improve model speed and user proximity [9] Market Dynamics & Competition - The AI sector is experiencing a technological revolution requiring significant infrastructure investment [4] - Demand for AI is rapidly growing, with OpenAI experiencing a tenfold increase in the last 18 months [3] - The US is currently ahead in AI spending compared to China [7] - Chinese companies like Tencent and BU are competing with global model makers like Anthropic and OpenAI [10] - Alibaba aims to become a more self-sufficient hyperscaler in the AI competition [10] Investment Risks & Rewards - The AI sector is expected to experience cycles of over and underinvestment, with potential for both gains and losses [5] - The long-term value of AI technology is projected to be significant for society [6] - Companies are being rewarded in the stock market for announcing substantial AI investments [12] - The risk of underinvesting in AI is currently perceived as more prominent than overinvesting [14]

Research Methodology & Focus Shift - The initial research aimed to predict naval movements and oil prices using satellite data but shifted to the effect of inflation on different income groups due to limited resources [5] - The research focus changed from inflation's impact on income groups to helping researchers conduct short-term research, using inflation as a case study [14][15] - The study incorporated mixed methods research, including quantitative analysis (regression models), thematic analysis of documents from government agencies, NGOs, and institutions like the World Bank, and surveys [12][13][16] Challenges & Solutions in Short-Term Research - Short-term research faces challenges as numbers may not have time to stabilize, leading to jittery models and patterns [10] - The researcher initially struggled with regression models that yielded already known information about inflation's impact [8] - To address the limitations of short-term data, the researcher supplemented quantitative analysis with thematic analysis of documents and surveys to humanize the research [12][13] Role of LLMs in Research - Large language models (LLMs) were used to determine the optimal sequence for mixed methods research, comparing quantitative-before-qualitative and qualitative-before-quantitative approaches [17][18][19] - The LLM analysis indicated that starting with quantitative analysis followed by qualitative analysis yielded richer interpretations [19] - The use of LLMs in research received mixed reactions from reviewers, with some seeing it as a red flag and others as a novel contribution [20][21] Key Takeaways & Recommendations - Researchers should be flexible and open to pivoting from their initial plans and hypotheses [24] - The research emphasizes the importance of open access and sharing knowledge in science [15] - The study suggests researchers should be open to learning and unlearning, actively researching in the field and discovering insights unexpectedly [25]

AI Fundamentals & History - AI has been studied for almost a century and integrated into daily life for decades, exemplified by facial recognition and voice assistants [2] - Large language models (LLMs) have driven recent AI advancements, making AI conversational and accessible [5] - AI systems learn from vast amounts of text and other data, enabling them to generate human-like text, but they lack human-level understanding, feelings, and consciousness [8] AI Risks & Ethical Considerations - AI-generated content raises copyright concerns due to the lack of mechanisms to trace the origin of training data and compensate original creators [12] - AI can perpetuate and amplify societal biases present in the data it is trained on, leading to discriminatory outcomes [19] - The use of AI for social scoring, as experimented with in some countries, raises concerns about privacy and restriction of personal freedoms [15] - The European Union's AI Act aims to regulate AI development and usage based on risk levels, prohibiting certain applications like social scoring [16] AI Limitations & Future Directions - AI systems, particularly LLMs, struggle with numerical and spatial reasoning [21][22] - It is crucial to educate and promote conscious development and usage of AI [24] - AI is not a magical solution but a tool that requires human intelligence to understand, regulate, and guide its development [25] - Research efforts, such as the EU-funded Infinity project, focus on improving the quality and representativeness of data used to train AI, particularly in the context of cultural heritage [20]

Industry Overview - Generative artificial intelligence (AI) is rapidly growing, with projections indicating it could become a $1.3 trillion market by 2032 [1] Company Profile: SoundHound AI - SoundHound AI aims to leverage the synergy between large language models and traditional speech recognition technologies to create voice AI platforms for various applications, including customer service and automotive assistants [3] - The company has established partnerships with notable restaurants like Krispy Kreme and White Castle, utilizing its software in drive-through locations [4] Value Proposition - SoundHound's automated systems can improve ordering time by approximately 10%, potentially increasing annual revenue by nearly $200,000 per location and enhancing profit margins by reducing labor costs [5] - Management claims that their systems are more accurate than human operators, which is crucial for customer satisfaction and long-term adoption in the industry [5] Performance Analysis - In the first quarter, SoundHound's revenue increased by 151% year over year to $29.1 million, although net losses rose slightly to $22.3 million [7] - The revenue growth has been significantly influenced by acquisitions, including the purchase of Allset and Amelia in 2024, complicating the assessment of the core business's performance [8] Financial Health - Despite revenue growth, SoundHound's gross margin decreased from approximately 60% to 37%, likely due to the integration of new acquisitions [10] - The company holds $246 million in cash with no debt, indicating it can sustain its current cash burn rate for the time being [10] Strategic Considerations - The challenge for SoundHound lies in effectively integrating its acquisitions into a cohesive business model over the next five years, which is often difficult to achieve [11] - Historical data suggests that 70% to 75% of acquisitions fail to meet their objectives, raising caution for investors regarding SoundHound's near-term growth and future prospects [12]

Core Company Insights - CoreWeave is an AI-focused cloud computing start-up that is experiencing significant growth due to increased spending in generative AI [1] - The company is benefiting from strong demand for GPUs, which are essential for training and running large language models [1] Market Trends - There is a massive upsurge in generative AI spending, which is driving the demand for CoreWeave's services [1]

Core Insights - Advanced Micro Devices (AMD) is experiencing significant growth in the data center market, primarily due to the adoption of its EPYC processors, with data center revenues increasing by 57.2% year over year to $3.674 billion in Q1 2025, representing 49.4% of total revenues [1][10] Group 1: Market Performance - The adoption of fifth-generation EPYC Turin processors and Instinct AI accelerators is driving growth, with hyperscalers like AWS, Google, and Oracle launching over 30 new EPYC-powered cloud instances [2] - Enterprise adoption of EPYC processors has doubled among Forbes 2000 customers in Q1 2025 [2] - EPYC processors are now utilized by all top 10 telecom, aerospace, and semiconductor companies, and are gaining traction in automotive, manufacturing, and energy sectors [3] Group 2: Technological Advancements - The integration of EPYC processors enhances performance per watt for containerized workloads, crucial for 5G Core, edge, and enterprise applications, thereby supporting telecom networks in meeting rising data demands while improving energy efficiency [4] Group 3: Competitive Landscape - AMD faces strong competition in the data center market from Intel and NVIDIA [5] - Intel is investing in Xeon processors and AI accelerators, with its Datacenter and AI Group revenues increasing by 8% year over year to $4.13 billion in Q1 2025 [6] - NVIDIA's data center revenues surged by 73.3% year over year to $39.1 billion in Q1 2026, driven by demand for generative AI and large language models [7] Group 4: Financial Performance and Valuation - AMD shares have gained 1.4% year to date, underperforming the broader Zacks Computer & Technology sector's return of 2.4% and the Zacks Computer - Integrated Systems industry's increase of 16.7% [8] - AMD's stock is trading at a forward 12-month Price/Sales ratio of 5.86X, compared to the industry's 3.59X, indicating a premium valuation [11] - The Zacks Consensus Estimate for Q2 2025 earnings is 56 cents per share, reflecting a 13.6% decline over the past 30 days, with a year-over-year decline of 18.84% [13]