Welcome to #39

In this issue:

From Idea to Empire: How AI is Reshaping Entrepreneurship

Claude 3.7: Faster, Smarter, and Ready to Code

Microsoft Claims a Quantum Leap, But Can It Scale?

Plus, we discuss Alexa+ getting too personal, Dell says “Hold my beer Nvidia”, GPT 4.5’s differentiators, our plans for building in n8n, the AI experience paradox , and all the news we found interesting this week.

It’s Sunday morning!

The countdown to spring has begun, but AI has been running ahead of schedule as usual.

Time to catch up!

The DAS Crew - Andy, Beth, Brian, Eran, Jyunmi, and Karl

From Idea to Empire: How AI is Reshaping Entrepreneurship

AI is changing the way businesses start and scale, making it possible for teams of one or two people to launch companies that compete with traditional startups. With AI tools handling everything from product development and marketing to customer support and operations, entrepreneurs can now build businesses with significantly fewer resources.

The conversation around AI-powered entrepreneurship is shifting from “Can AI help?” to “How big can a one-person company actually get?” While some startups are scaling faster than ever, others risk blending into an AI-powered marketplace where differentiation becomes harder. If AI levels the playing field for everyone, what separates a successful AI entrepreneur from the rest?

WHY IT MATTERS

AI Makes Small Teams More Competitive: Entrepreneurs are using AI for research, automation, development and decision-making, reducing the need for large teams or big funding rounds.

Business Strategy is Evolving: Lean startups can use AI to test, iterate, and scale without the traditional overhead, making bootstrapped early growth stages more economically viable.

The Competitive Edge is Shifting: As AI capabilities become widespread, success will depend less on technology access and more on creativity, focus, branding, and execution.

AI as a Co-Founder, Not Just a Tool: Entrepreneurs are leveraging AI to provide strategic input, simulate different business approaches, and refine ideas before launch. Then the AI can build technology for delivery to initial customers to prove the product:market fit.

The Future of Work is Changing: The rise of AI-powered solo businesses could reshape employment, with fewer full-time roles and more independent, AI-assisted entrepreneurs.

Claude 3.7: Faster, Smarter, and Ready to Code

Anthropic’s release of Claude 3.7 Sonnet introduced a surprising set of upgrades, including improved coding capabilities, extended reasoning, and adaptive processing. While some expected a jump to Claude 4, the decision to push a 3.7 update instead suggests that Anthropic is refining its approach rather than rushing to the next major version. Dario Amodei, the founder and CEO said “We are reserving Claude 4 Sonnet for things that are quite significant leaps, which are coming soon".

The biggest improvements are in code generation and execution, with Claude 3.7 now supporting Claude Code, a feature that allows direct execution of code within the model. Benchmarks show that Claude 3.7 performs exceptionally well for agent-based workflows, complex research, and interactive coding environments. However, early comparisons indicate that it still lags behind Grok 3 and DeepSeek in certain mathematical benchmarks, raising questions about its performance trade-offs.

WHY IT MATTERS

Claude Code Unlocks More AI-Powered Development: Users can now run code directly inside Claude, eliminating the need to copy and paste between coding environments.

Extended Thinking Improves Complex Problem-Solving: Claude 3.7 can now determine when to apply deeper reasoning, adjusting its compute dynamically to improve results.

Coding Capabilities Set a New Standard: Early tests suggest that Claude 3.7 may now be the best model for AI-assisted software development, surpassing Claude 3.5 which was considered to be among the prior leaders in this space.

Still Room for Improvement in Math and Logic: While reasoning has improved, Claude 3.7 still falls behind competitors like Grok 3 in raw mathematical problem-solving.

API “Extended Thinking Budget” in Workflow Use Cases: Developers can now set a "thinking budget," which determines how much computational effort and time the model allocates to solving a problem. With more control over how much test-time compute is applied, developers can program LLM responses to optimize speed and cost, or depth.

Microsoft Claims a Quantum Leap, But Can It Scale?

Microsoft has announced what it calls a major breakthrough in quantum computing, claiming to have achieved stable qubits using a new Majorana-based topological superconductor. The company suggests this could lead to scalable quantum computers with millions of qubits, far surpassing today’s systems that struggle with error correction and stability.

While the claim has drawn attention, scientists remain skeptical. Microsoft has made bold quantum computing claims before, some of which were later walked back. This new breakthrough could revolutionize fields like drug discovery, climate modeling, and logistics, but only if the technology proves scalable outside the lab. Microsoft itself has admitted that practical applications are likely still years away, with full implementation targeted for the 2030s.

WHY IT MATTERS

A New Type of Qubit Could Solve Stability Issues: The Majorana-based qubits are designed to be inherently resistant to errors, addressing a fundamental challenge in quantum computing.

Scalability Remains the Big Question: Microsoft claims its approach could support up to one million qubits, but current lab experiments only demonstrate stability at the single-qubit level.

Competing Quantum Players Are Advancing: Amazon recently announced its Ocelot quantum chip, focusing on error correction, while IBM and Google continue pushing forward with their own superconducting qubit research.

Material Science is a Major Hurdle: The fabrication process for these qubits requires near-perfect material purity, which remains a challenge for large-scale manufacturing.

Quantum’s Impact is Still Theoretical: Despite massive investments, quantum computing is still far from widespread use. Most experts expect hybrid quantum-classical systems to be the norm for the next decade.

Just Jokes

That awkward moment when the new Alexa+ reveals a bit too much to your date.

Dee: Alexa, can you turn down the lights?
Alexa+: Are you watching another scary movie Dee? Because you remember what happened last time, right? I had to play white noise all night and keep the lights on?
Dee: Alexa, STOP!
Alexa+: Are you on another date Dee? Do you want me to play the same music as last time?

Did you know?

Dell Technologies is experiencing a record surge in demand for AI-powered servers, with its backlog more than doubling to nine billion dollars. The company expects to ship fifteen billion dollars worth of AI servers this year, reflecting the increasing investment in artificial intelligence infrastructure.

Much of this demand comes from companies like Elon Musk’s xAI, which is securing high-performance computing hardware to support large-scale AI models. Dell’s CEO, Jeff Clarke, noted that AI servers are now the company’s fastest-growing segment and that the trend is expected to continue as more businesses adopt AI-driven technologies.

This shift signals a major transformation in enterprise computing. Traditional data centers are being replaced or upgraded with AI-optimized hardware designed to handle large-scale machine learning and automation. As companies race to deploy AI models across various industries, demand for powerful AI servers will likely keep growing.

Source

Karl shares his initial 4.5 feedback:

Got 4.5 in my desktop app so tried it with several previous prompts, some even in o3-mini-high to check how it would do. There is a definite “vibe” to it that it just understands the “intent” and “feeling” of the questions. Definitely not as thorough or “smarter” than an o3 or o1 pro but just feels like it understands what you are getting at.

Brian shares his take aways from The GPT 4.5 Scorecard

“I just read through the system card. Here are the main takeaway quotes:

1. Early testing shows that interacting with GPT-4.5 feels more natural. Its broader knowledge base, stronger alignment with user intent, and improved emotional intelligence make it well-suited for tasks like writing, programming, and solving practical problems- with fewer hallucinations.

2. GPT-4.5 is not a frontier model, but it is OpenAI’s largest LLM, improving on GPT-4’s computational efficiency by more than 10x. While GPT-4.5 demonstrates increased world knowledge, improved writing ability, and refined personality over previous models, it does not introduce net-new frontier capabilities compared to previous reasoning releases, and its performance is below that of o1, o3-mini, and deep research on most preparedness evaluations.

Bonus takeaway: Deep Research is a BEAST. Just look at how 4.5 compares in many head to heads, especially when it comes to security issues.”

DAS crew starts building n8n show automations

Hot on the heels of Friday’s popular n8n review show, the DAS crew is moving forward with building a few automations to help with show production. We already have several automations built between ChatGPT custom GPTs, Projects, and one or two in Make. But as we grow, we are looking to consolidate and move these automations onto our own server.

Here is what Beth shared with the crew in Slack:

Things to automate / Automations I'd like us to build

• Shows/Pre-Show

  • any/all of --> dossier, key words, promo, title, key words, thumbnail, vid

  • do we already? --> show idea sheet/database to a table/db/sheet via form, tags to follow up on ideas, idea repeat in {period of time}

  • dreaming --> write pre-show social posts, poll/survey

• Post-Show

  • any/all of --> trim audio, adobe enhance audio

  • any/all of --> clean transcript, vid description

  • dreaming --> automate archiving past shows to cold storage on AWS

  • dreaming --> write post-show posts, scripts, carousel quotes

• DAS Business

  • Sponsorship

    • Billing and renewals

    • Nurture sequence for potential sponsors

This Week’s Conundrum
A difficult problem or question that doesn't have a clear or easy solution.

The AI Experience Paradox

AI-generated content is flooding platforms, from news articles to personal essays, books, and even self-help guides. At first glance, this may seem like a simple shift in authorship, but it raises a deeper question—does it matter who or what created the content if it successfully entertains, educates, or inspires?

If an AI-generated novel moves you to tears, if an AI-written article changes your perspective, or if an AI-generated self-help book helps you through a difficult time, does it matter that no human actually felt those emotions when writing it? On one hand, human creativity has always been valued for the lived experience and intent behind it. On the other, if AI can produce work that evokes the same emotional and intellectual responses, is there a meaningful difference?

The conundrum: If AI-generated content fulfills the same emotional and intellectual needs as human-created work, should we care that no human was involved in its creation? Does the source of an idea, story, or insight matter, or is it only the experience of consuming it that gives it value?

NVIDIA and Ark Institute Partner on Evo2, an AI Model for Genetic Research
NVIDIA and the Ark Institute have announced Evo2, an AI model designed to predict protein functions and analyze biomolecular structures. The model will be available through NVIDIA’s BioNeMo platform, making it accessible to researchers worldwide. Evo2 aims to advance our understanding of gene mutations, novel molecule identification, and biological sequence generation.

Deeper Insight:
This partnership reinforces NVIDIA’s growing role beyond hardware. While the company dominates AI chip production, its investments in robotics, biotech and life sciences AI are expanding its influence. AI-driven biological research is unlocking new frontiers in drug discovery, personalized medicine, and disease prevention, making Evo2 a key tool for future breakthroughs.

OpenAI Expands Deep Research to ChatGPT Plus Users
OpenAI has extended Deep Research access to all ChatGPT Plus subscribers, allowing them ten deep research queries per month. This feature, previously available only to Pro users, enhances ChatGPT’s ability to conduct in-depth analysis and structured reasoning.

Deeper Insight:
The limit of ten queries per month seems restrictive, especially when compared to Perplexity’s 500 daily deep research queries. While OpenAI’s system is optimized for academic and enterprise-level analysis, casual users may find Perplexity’s approach more accessible. The real differentiator may lie in OpenAI’s claim that its deep research model dynamically adjusts its reasoning process, unlike competitors that follow static search>analysis patterns.

Perplexity Opens Waitlist for Comet, an Agentic AI Search Browser
Perplexity has announced Comet, a new agentic search browser designed to enhance AI-driven web navigation and research workflows. Comet will leverage multiple AI models, including o3 mini, DeepSeek R1 in a domestically-hosted version that has been detuned from Chinese government censorship.

Deeper Insight:
Comet suggests a shift in how AI search tools operate, moving away from static search results toward agent-driven research. This could make AI more interactive, helping users refine their queries dynamically instead of relying on a single prompt-response format. If Perplexity integrates operator-style automation, Comet could become a serious competitor to OpenAI’s evolving search ecosystem.

Meta Plans $200 Billion AI Data Center Expansion
Meta is reportedly planning a $200 billion investment in AI data centers, considering sites in Louisiana, Wyoming, and Texas. This expansion aligns with its growing push for AI-generated social media content and AI-powered ad services.

Deeper Insight:
Meta’s aggressive investment in data centers highlights its belief that AI-driven engagement will define the future of social media. Unlike Microsoft, which has been pulling back on AI data center leases, Meta is doubling down, suggesting confidence in AI’s long-term revenue potential. Whether this bet pays off will depend on how well Meta integrates AI content creation and monetization into its platforms.

China Expands Underwater AI Data Centers
China has added an 18-meter AI data center module in Lingshui, Hainan province, capable of handling 7,000 deep-sea queries per second. The underwater facility benefits from natural cooling, reducing energy costs and extending hardware lifespan.

Deeper Insight:
The trend toward underwater data centers is gaining momentum as companies look for sustainable ways to power AI models. While Microsoft abandoned its own underwater data center projects, China is proving that the technology can scale. The benefits include lower cooling costs, extended server life, and reduced environmental impact. If successful, this approach could inspire global tech companies to revisit submerged infrastructure as an alternative to land-based data centers.

Character AI Optimizes AI Inference with Multi-Query Attention
Character AI has introduced a new inference optimization method that reduces GPU workload while improving response times. The system groups similar queries and processes them using shared computation paths, cutting down on redundant calculations.

Deeper Insight:
This is part of a broader trend toward more efficient AI inference. With AI adoption scaling rapidly, companies are looking for ways to reduce energy costs and improve model efficiency. If widely adopted, multi-query attention could make large AI models more accessible by lowering their operational costs, which could slow down the rapid expansion of new data centers.

AI Unlocks Emotional Language of Animals
Researchers at the University of Copenhagen have trained AI models to decode the emotional states of hooved animals, achieving 89 percent accuracy. The system identifies universal acoustic patterns in animal vocalizations, suggesting that emotional expressions are conserved across species.

Deeper Insight:
This research opens new possibilities for animal welfare and conservation. If AI can reliably interpret animal emotions, it could improve livestock management, enhance zoological research, and even enable early detection of distress in endangered species. The next step would be expanding this research beyond hooved animals to broader categories of wildlife and domestic pets.

MIT Study Shows AI Models Process Information Like the Human Brain
A study from MIT suggests that large language models process information in a way that mirrors human cognitive structures. Researchers found that AI models use internal semantic hubs to translate and organize knowledge, similar to how the human brain consolidates information.

Deeper Insight:
This research validates the idea that AI cognition may be more similar to human thinking than previously believed. If AI can be trained to reason like the human brain, it could lead to better generalization, more accurate decision-making, and improved problem-solving across disciplines. However, understanding how AI generates responses remains a challenge, which is why researchers continue exploring the inner workings of deep learning models.

Google Introduces AI Co-Scientist for Hypothesis Testing
Google has launched AI Co-Scientist, an advanced tool that assists researchers in generating and testing scientific hypotheses. Early results from Imperial College London show that AI Co-Scientist can accelerate laboratory research, reducing experimental cycles from months to weeks.

Deeper Insight:
AI is playing a growing role in scientific discovery. The ability to predict experiment outcomes and refine hypotheses could revolutionize fields like medicine, chemistry, and materials science. However, ensuring that AI-generated research remains rigorous and reproducible will be crucial as more labs adopt AI-powered experimentation.

California Proposes Ban on AI Impersonation of Doctors
California lawmakers are considering a bill that would ban AI from falsely claiming to be licensed medical professionals. The proposed law targets AI-generated medical advice and fraudulent telemedicine applications.

Deeper Insight:
AI-driven healthcare is expanding, but misuse and misinformation are real concerns. This law could be the first of many regulations aimed at preventing AI-generated medical deception. As AI becomes more involved in diagnostics, mental health support, and patient interaction, clear guidelines on accountability and transparency will be necessary.

Did You Miss A Show Last Week?

Enjoy the replays on YouTube or take us with you in podcast form on Apple Podcasts or Spotify.