Contrary Research Rundown #127
OpenAI goes to Washington (again), plus new memos on CloudKitchens, Nova Credit, and more
Contrary is hosting our first Tech Talk of the year in SF with Index Ventures, featuring eng leads and founders from Baseten, Codeium, Mercor, Patch, and Doss.
An evening built by engineers for engineers — each company will live demo their latest product innovations for leading builders in the Bay Area. RSVP here!
Join Contrary and Ramp in NYC on April 3rd for Undo. Undo. Redo. - a celebration of both design failures and triumphs. This lightning-talk style event for senior designers will feature unabridged stories from Contrary, Ramp, Anthropic, and Notion.
Register here for the chance to join!
Research Rundown
This Thursday, OpenAI submitted a policy proposal to the US government that has drawn a lot of attention and generated a lot of clickbaity headlines. Many journalists focused on the fact that the policy proposal directly attacks DeepSeek, an OpenAI competitor.
But the real points people should be taking away from the proposal are more nuanced. This isn’t just a case of OpenAI attacking a competitor or trying to erect favorable regulatory moats around its business (although that could certainly be one of the underlying motivations of the proposal). Instead, the proposal speaks to the crossroads the US, and the world, faces with AI.
It originated as a response to the White House’s call for public comment on an AI Action Plan that the government is putting together as the result of an executive order issued in late January aimed at “removing barriers to American leadership in artificial intelligence”, which called on the Office of Science and Technology Policy to submit an action plan to the President within 180 days. OpenAI’s proposal had several key points:
AI progress is accelerating. OpenAI points to the fact that “the cost to use AI falls by 10x every 12 months”, that “the amount of calendar time it takes to improve an AI model keeps decreasing”, and “AI models are catching up with human intelligence at an increasing rate.” It also stated that it believes we are “approaching artificial general intelligence”.
AI should be shaped by democratic principles. OpenAI argues that we must ensure “freedom of intelligence” ****by guarding against “autocratic powers that would take people’s freedoms away” and the “layers of laws and bureaucracy” that would prevent AI progress in democratic countries.
China is “determined to overtake us by 2030”. OpenAI says that in order to prevent China from overtaking the US, the US needs to do a number of key things: “export democratic AI”, increase partnership between the private sector and the federal government on AI, remove regulatory barriers around copyright for AI training data, increase investment in AI infrastructure, export “democratic AI” and promote the global adoption of American AI systems, and ban Chinese AI companies like DeepSeek.
Throughout the proposal, there’s an emphasis on the urgent threat posed by China. OpenAI paints a dark picture of what it calls Chinese “authoritarian AI”, which the CCP can use as a tool “to amass power and control their citizens”. It also makes the argument that China’s authoritarian regime gives it a number of “strategic advantages”, including the ability to quickly marshal resources to subsidize its AI industry, immunity to western IP laws which frees up its ability to use data to train its models, and “regulatory arbitrage” being created by the 781 state-level AI bills introduced in 2025 alone that could dampen American AI innovation.
In particular, OpenAI singles out DeepSeek as representative of “authoritarian AI” and its risks. The proposal says that “DeepSeek is simultaneously state-subsidized, state-controlled, and freely available, the cost to its users is their privacy and security” and that “[DeepSeek’s] models also more willingly generate how-to’s for illicit and harmful activities such as identity fraud and intellectual property theft, a reflection of how the CCP views violations of American IP rights as a feature, not a flaw.”
OpenAI also repeatedly mentioned Huawei in the same breadth as DeepSeek, clearly alluding to the precedent set by the US government’s numerous bans on the Chinese telecom company due to national security concerns. These bans began in 2017 and escalated in scope through 2022, spanning both the Trump and Biden administrations. OpenAI’s intent in mentioning Huawei so frequently was clearly to suggest similar actions should be taken on DeepSeek.
OpenAI also suggested a number of so-called “freedom-focused policy proposals” to countermand China’s AI advantages and “advance democratic AI”:
First, it proposed a framework for “voluntary partnership between the federal government and the private sector” on AI.
Second, it proposed export controls targeted at restricting “the flow of AI technologies to the PRC”, with a three-tiered system to govern AI diffusion.
Thirdly, it argued that fair use copyright laws should apply to AI companies using copyrighted data to train their data, arguing that “unpredictable availability of inputs hinders AI innovation, particularly for smaller, newer entrants with limited budgets.”
Finally, it said that the US should increase investment in AI infrastructure, arguing that “hundreds of billions of dollar sin global funds are waiting to be invested in AI infrastructure” and that if the US doesn’t channel those resources into its own projects, “the funds will flow to projects backed and shaped by the CCP”.
Some may point to the apparent contradiction in OpenAI simultaneously calling for export restrictions and bans on foreign competitors like DeepSeek (i.e. greater regulation and government involvement) while also arguing for the removal of regulatory barriers like copyright laws as a case of wanting to have their cake and eating it too.
However, the policy proposal makes a compelling case that US national security interests are aligned with OpenAI’s recommendations. Irrespective of OpenAI’s ulterior motives for the message it’s sending, it is choosing the right time to send it and the right government to send it to. The Overton window around protectionism and free trade has been smashed wide open by the new administration in recent months, and by positioning its policy proposal as a battle plan for the widening AI front in the escalating trade wars surrounding the US, it wouldn’t be surprising to see many of its recommendations echoed in the final version of the AI Action Plan that the Trump Administration ultimately adopts come July.
Nova Credit has built infrastructure to translate international credit data into a standardized format that domestic financial institutions can understand. To learn more, read our full memo here and check out some open roles below:
Senior Product Manager - New York, NY
Senior Solutions Architect - New York, NY
CloudKitchens is working toward a fully autonomous, hyper-efficient food ecosystem, positioning itself as the AWS of food infrastructure - a backend system that powers digital-first restaurant brands worldwide. To learn more, read our full memo here and check out some open roles below:
Backend Software Engineer - San Francisco/Bay Area, CA
Full-Stack Software Engineer - Mountain View, CA
Check out some standout roles from this week.
Maven Clinic | New York, NY or Remote (US) - Senior Software Engineer (AI/ML), Senior Software Engineer (iOS), Senior Product Designer, Senior Director of Product (Maternity Program), GTM (Systems Lead)
Codeium | Mountain View, CA - Deployed Engineer (Pre Sales), Security Operation Engineer, AI Product Engineer, Software Engineer (High Performance ML)
Navan | Palo Alto, CA - Back-End Engineer, Senior iOS Engineer (Travel Team), Senior Back-End Engineer (AI), Lead Artificial Intelligence (AI) Researcher, Lead Product Manager (Fraud)
Klarna is restructuring teams, consolidating roles, and cutting costs to become leaner ahead of its potential IPO, leaving its employees in a state of uncertainty.
Allbirds, once a high-flying IPO, has seen its stock price plummet over 60% in just a year and has a current market cap of just over $41 million.
Stanford students are increasingly interested in working on military and defense technology, seeing it as a more "mission-driven" and impactful career path compared to traditional tech jobs.
Anthropic CEO Dario Amodei is concerned that Chinese spies are stealing valuable "algorithmic secrets" worth $100 million from top U.S. AI companies in just a "few lines of code", and he wants the U.S. government to step in and help defend against this threat.
Intel's new CEO Lip-Bu Tan is widely respected in the tech industry, serving on the boards of companies like Hewlett Packard and SoftBank, and is expected to help drive Intel's renewed focus on customers and delivering for shareholders.
Redpoint Ventures published a market overview presentation providing insights into the current state of the public and private technology markets.
Cartesia, a voice AI startup, has raised a $64 million Series A led by Kleiner Perkins to develop real-time generative AI models that reduce computational costs and enable large-scale data handling.
Scopely, the makers of the mobile game Monopoly Go, have acquired Pokémon Go and Niantic’s entire games business for $3.5 billion.
Nvidia invested $100 million in Coreweave and committed to spending $300M/year renting back their GPUs.
Amazon, Google, and Meta have pledged to support tripling nuclear power capacity by 2050.
Substack has seen a surge in paid subscriptions, reaching over 5 million, driven in part by high-profile creators leaving mainstream media outlets like CNN and MSNBC to launch their own Substack newsletters.
Waymo is beginning public autonomous vehicle service in the San Francisco Peninsula area, starting with Palo Alto, Mountain View, and Los Altos.
Lovable's growth increased by almost 50% last week, now adding 1,500 customers per day.
Hinge Health, a digital physical therapy provider, filed for an IPO. The company’s revenue has grown from $293 million in 2023 to $390 million in 2024, while its net loss shrank from $108 million to $11.9 million over the same period.
CoreWeave has signed a $11.9 billion, five-year deal with OpenAI, which will receive a $350 million stake in the AI data center provider ahead of CoreWeave's expected IPO.
Eric Schmidt, the former CEO of Google, is taking over as the CEO of Relativity Space, a 9-year-old rocket startup, marking his first CEO role since leaving Google nearly 15 years ago.
Rocket Companies, a $31 billion finance and real estate holding firm, is acquiring Redfin for $1.75 billion in an all-stock transaction.
At Contrary Research, we’ve built the best starting place to understand private tech companies. We can't do it alone, nor would we want to. We focus on bringing together a variety of different perspectives.
That's why applications are open for our Research Fellowship. In the past, we've worked with software engineers, product managers, investors, and more. If you're interested in researching and writing about tech companies, apply here!