Welcome back. OpenAI just landed a $200 million DoD contract to develop AI for "warfighting and enterprise domains"—officially launching their new "OpenAI for Government" division. Remember when they had policies against military use? Apparently that was so last year, and $200 million is excellent motivation for a policy pivot.

— The Deep View Crew

In today’s newsletter:

• 💬 AI for Good: Teaching AI to care by making medical chatbots more human

• 📅 Microsoft and OpenAI talks hit eighth month with tensions rising

• 👀 Forget the past, AI investors have eyes on the future

💬 AI for Good: Teaching AI to care by making medical chatbots more human

When patients reach out to chatbots with scary symptoms, they're often dealing with more than just physical concerns. They're anxious, frightened and looking for reassurance alongside medical advice.

Researchers at National Taiwan University figured this was a problem worth solving.

Here's how they did it: The team took real doctor-patient conversations and rewrote them to include patient messages expressing fear, anxiety, embarrassment, frustration and distrust. Then they crafted doctor responses designed to provide both accurate medical information and emotional comfort.

• Using this modified dataset, they fine-tuned Llama language models with three different training methods.

• The approach that worked best — called Direct Preference Optimization — significantly improved the models' ability to deliver empathetic responses while maintaining medical accuracy.

The results: Models trained on the emotional data consistently outperformed standard medical chatbots across empathy metrics. When patients expressed fear about symptoms, the upgraded AI could respond with phrases like "It's completely understandable to feel concerned" while still providing solid medical guidance.

This research highlights a gap that shouldn't exist in the first place. The fact that medical AI systems need special training to show basic human empathy reveals how far we still have to go in making these tools truly helpful rather than just technically correct.

Still, for patients stuck with AI-powered telehealth platforms — which is increasingly common — chatbots that can balance knowledge with compassion represent a meaningful step forward.

Orchestrating LLM workflows with the Airflow AI SDK

Most data teams need reliable LLM workflows that solve real business problems, not agents talking to agents.

In this webinar, you’ll learn how to use open-source AI SDK for Apache Airflow, built to simplify the creation of production-ready LLM workflows.

Register now to join Astronomer on June 26 to learn from Julian LaNeve, Astronomer CTO and creator of the AI SDK, as he walks through:

• Key differences between LLM workflows and agents and when to use each

• How to integrate LLMs into your Airflow pipelines to drive quicker results with less complexity

• Live demo of the AI SDK in action and see how it fits into the Airflow framework

Save your spot today.

📅 Microsoft and OpenAI talks hit eighth month with tensions rising

Microsoft and OpenAI are locked in increasingly tense negotiations after eight months of talks, with OpenAI executives reportedly considering a "nuclear option" of filing federal antitrust complaints against their biggest partner and investor.

The conflict centers on OpenAI's planned $3 billion acquisition of coding startup Windsurf, which directly competes with Microsoft's GitHub Copilot. Under current agreements, Microsoft's $13 billion investment grants it access to all OpenAI technology, including acquisitions. OpenAI wants to block Microsoft from accessing Windsurf's intellectual property, creating what sources describe as a "standoff."

OpenAI faces a December 2025 deadline to restructure as a public benefit corporation or risk losing $20 billion in funding from SoftBank. The company wants Microsoft to accept a 33% equity stake in exchange for waiving future profit rights, but Microsoft seeks additional protections for its investment.

Currently, Microsoft receives 20% of OpenAI's revenue through 2030 and maintains exclusive hosting rights. OpenAI has already ended Microsoft's cloud exclusivity, partnering with Google Cloud and Oracle, and wants to reduce Microsoft's revenue share to 10%.

The "nuclear option" involves OpenAI accusing Microsoft of anticompetitive behavior to federal regulators. This comes as the FTC already investigates their partnership for potential antitrust violations, with the previous Chair warning about partnerships that "create lock-in" and "stifle competition."

Microsoft CEO Satya Nadella and OpenAI's Sam Altman, who previously texted daily, now communicate through scheduled weekly calls as relations have cooled.

With OpenAI generating $10 billion in annual revenue and Microsoft's AI business approaching similar figures, the outcome could reshape how tech giants structure AI alliances and whether regulators impose new restrictions on such partnerships.

Most car factories, like Ford or Tesla, reportedly build one car per minute. Isn't it time we do that for houses?

BOXABL believes they have the potential to disrupt a massive and outdated trillion-dollar building construction market by bringing assembly line automation to the home industry.

BOXABL homes are built in their Las Vegas factory, folded, shipped on a truck, and then unfolded on site in one hour. They aim to transform home construction, much like how Henry Ford automated car manufacturing – bringing assembly line efficiency and mass production to an artisanal, slow industry.

And they're not just dreaming big; they're delivering:

• Initial prototype order delivered to SpaceX in 2020.

•  Subsequent project order of 156 homes from the Department of Defense completed in 2021.

• Now, after implementing what was learned from those prior orders, actively delivering to developers and consumers.

• BOXABL reserved Nasdaq ticker symbol $BXBL*!

BOXABL has already raised over $200M from 50,000+ investors since 2020, and recently achieved a significant milestone: raising over 50% of their Reg A+ funding limit!

Like other game-changing companies, you have a chance to invest in BOXABL’s offering at just $0.80/share. All BOXABL crowdfunding will close on June 24th. This includes all accredited and non-accredited offerings.

Invest now before it’s too late!

This is the LAST CHANCE to invest in the current BOXABL funding round and claim your bonus shares.

• Amazon plans more job cuts as AI takes over

• G7 leaders want AI running more government work

• WhatsApp is finally getting ads whether you like it or not

• Top Trump labor official: America's workers don't trust employers with AI

• Coralogix rolls out a new AI agent after landing $115 million

• A CEO playbook for turning gen AI into real business results

• Meta: Software Engineer, Machine Learning

• AMD: Machine Learning Software Engineer (SMTS)  

• Humata: Query files and let AI summarize findings

• Glasp: Highlight and save web content, then summarize it with AI for later use

• Granola: AI notetaker Granola is now available on Windows

👀 Forget the past, AI investors have eyes on the future

The AI funding boom shows no signs of slowing. Despite a wave of disappointing outcomes from well-funded startups like Character.ai and Inflection AI, venture capitalists are moving forward with even larger bets. 

By the numbers

• Over the past year, they have poured $52.4 billion into generative AI companies—surpassing the $32 billion total invested between 2022 and mid-2024.

• The average deal size jumped from $96 million to $372 million as investors moved from sprinkling capital across dozens of experiments to concentrating resources on perceived category winners.

• SoftBank and Thrive Capital now top the AI investor rankings by total deal value, accounting for more than $20 billion of recent funding. Neither firm ranked in the top nine just one year ago.

  • Both led multiple OpenAI rounds and purchased shares from employees and early investors.

  • Thrive also backed infrastructure plays like its $900 million investment in coding assistant developer Anysphere, valued near $10 billion, and led a $600 million round for Alphabet's (Google) AI drug discovery unit Isomorphic Labs.

Traditional West Coast firms remain active but have shifted their approach. Lightspeed led 11 deals worth nearly $4 billion over the past year. Andreessen Horowitz led 22 rounds in the same period, including repeat investments in ElevenLabs and early support for Mistral AI and Character.ai. Accel is expected to see a multibillion-dollar return from Meta’s $14.3 billion investment in Scale AI, where it is the largest outside investor.

Andreessen returned to lead multiple ElevenLabs rounds, including one at a $3.3 billion valuation in January. The follow-on investments expose firms to bigger wins—but also more concentrated risk.

Core AI developers have dominated fundraising. OpenAI raised $6.6 billion from Thrive last fall and $10 billion from SoftBank in April, with plans for a $40 billion round later this year. Anthropic secured $3.5 billion from Lightspeed. Greenoaks invested $2 billion in Safe Superintelligence (SSI), launched by OpenAI's former chief scientist. Even newer entrants like Musk's xAI attracted $6 billion.

Yes, but: The flood of capital has encouraged a new wave of AI startups, but it has also intensified competition in crowded categories. Founders building agents, model evaluators and workflow automation tools are facing saturation and investor fatigue. Many cannot raise follow-on rounds without major differentiation or proven traction. 

That has led to a bifurcation in the market. Top-tier technical teams continue to attract capital at rising valuations. Others struggle to stay afloat. The gap between the leaders and everyone else is widening.

Since 2022, 724 funding rounds across 507 generative AI startups have raised more than $85 billion. The top nine investors alone led 74 rounds worth $27.5 billion in just the past year. Accel alone has closed another $2.68 billion in deals that remain unannounced.

Venture capital's "spray and pray" era is ending. The biggest AI investors are placing concentrated bets on perceived market leaders—primarily core infrastructure and foundational models.

Anysphere leapt from an $8 million seed round in October 2023 to $900 million just 20 months later. Meanwhile, overall seed funding collapsed to $13.2 billion in 2024—one-third below the 2022 peak.

This is venture gentrification—big money crowding out the garage labs that drive paradigm shifts. The focus sounds prudent, but it's profoundly conservative. DeepMind, Stable Diffusion and Whisper all began on shoestring budgets; a world that funds only giants may never see their equivalents.

We're trading the messy, inefficient process that produces genuine surprises for neat efficiency that produces predictable consolidation. In a field where the next breakthrough could come from anywhere, we're systematically defunding "anywhere" in favor of "here."

🤔 Your thought process:

Selected Image 1 (Left):

• “The shadows all seem right and the throw on the couch is just the right amount of messy. The other image has weird stuff in the mirror.”

• “The title on the book was the big one for me as well as the shadows.”

Selected Image 2 (Right):

• “The crown molding didn't meet properly in the corner and the door frames were mismatched in the first image. Also, the reflection in the mirror seemed incorrect.”

• “Impressive! The reflection in the mirror had me sold...”

💭 A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

P.S. Enjoyed reading? Take The Deep View with you on the go! We’ve got exclusive, in-depth interviews for you on The Deep View: Conversations podcast every Tuesday morning. Subscribe here!

If you want to get in front of an audience of 450,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

^{*Indicates sponsored content}

^{*Boxabl Disclosure: This is a paid advertisement for BOXABL’s Regulation A offering. Please read the offering circular here. This is a message from BOXABL}

^{*Reserving a Nasdaq ticker does not guarantee a future listing on Nasdaq or indicate that BOXABL meets any of Nasdaq's listing criteria to do so.}