Cal Café — Berkeley Founders, Funders, and Friends, hosted by Canvas Prime & Accel. This Thursday, March 26 at the Accel office in South Park. Sign up here. ☕🐻
Deal of the Week
Congrats to Imperative Care and their $100M raised to continue the commercialisation of their stroke and vasular thrombectomy portfolios. Looks a really interesting company that will make lives better!
Design Philosophy in the Age of AGI
It seems each week the models are getting smarter and converging on each other in “intelligence,” but diverging in design philosophy.
If you jump between Claude and ChatGPT, it does feel like ChatGPT is increasingly nudging you—giving you the answer, but then ending with an “If you want, I can…” or trying to pull you into the next query.
Part of this might be the Meta DNA permeating through OpenAI. In October, The Information estimated that ~20% of OpenAI employees were ex-Meta and they just announced another big hire today from Meta. That’s a meaningful number—and perhaps not surprising if they’re trying to replicate the success of one of the most powerful engagement engines ever built.
Sometimes I like this, but increasingly it’s turning me off.
Claude behaves differently. It’s more restrained and doesn’t seem to nudge as much. That’s clearly a design decision, but it’s increasingly interesting how these editorial choices are diverging. Potentially driven by different business and growth priorities.
All of this takes me back to an idea that was discussed at a founder dinner recently. Will all front-end UX be vibe-coded? It feels like it might be possible to do this, but does that make it probably... I think distinct design philosophy will still matter, because most users don’t know exactly what they want. But they know something great when they see it. The second-order effect here may be that the price you can charge for software comes down.
The models may end up looking the same. But how they choose to behave won’t.
Only the Phoenixes survive in the AI era.
Something that is playing out writ large across enterprises and the public markets is two things. First, a proliferation of generalized AI agents companies automating workflows, and second, models getting better and better at many of these tasks (see my old post on SaaS-apocalypse). I increasingly believe this new tech won’t wipe out legacy SaaS companies, but it has fundamentally changed the discount rate for them and will likely force a reduction in stock-based compensation. I digress!
The more interesting shift is strategic. A $5B+ second time founder I know scrapped a year of real traction on a horizontal AI agent to go fully vertical. That felt extreme. Until you realize the horizontal layer is commoditizing in real time. It made me think of a Phoenix, the mythical bird that can regenerate and be born again.
You’re seeing versions of this everywhere. Replit was initially tooling for developers, and now they’ve reinvented themselves as a vibe coding app and just hit a $9B valuation. Intercom, an older-generation customer support solution that was flatlining, saw its founder return as CEO and launch Fin, their AI offering (there’s a great article on how hard this was from the founder).
A great example from the Canvas Prime portfolio: Casetext (legal AI) gained early access to the magic of LLMs through GPT-4 and was able to pivot all its efforts toward this, resulting in a $650M exit in 2023 to Thomson Reuters. Thomson Reuters just announced one million users for the Casetext CoCounsel product. Also a Berkeley founded company through Pablo Arredondo.
The takeaway: survival increasingly requires “Phoenix mode”. A willingness to burn what’s working and rebuild around where the world is going, not where it is.
Quick Takes
Back in September 2025 I posted this:
Application Companies to Creep into the Model Layer?
A really interesting article by Berkeley alum Surya Dantuluri. He posits that some AI applications should and do train their own models, or sell user actions back to the labs.
For example, Cursor can use RL from customer actions to build better models for specific features, allowing them to serve at a lower cost compared to paying for an API from the latest frontier model.
What are the long-term impacts of this? It can either be less favorable for the big model companies as their frontier models are only used (and paid for) in certain instances or you think that the total TAM expands so much that it ultimately doesn’t matter. That might well be the case.
Then in October 2025 this idea - ‘The Atomisation of Data pipeline a big threat to the Foundational model providers?’ .
It seems I was early on this and Cursor just announced their own model which appears to be a tuned up version of the Kimi 2.5 Opensource model. The lines are increasingly blurring between the the large application and the models!
Summary by the #️⃣ & 💰:
3 Berkeley-founded companies funded
$105M of capital raised from the 16th March to 22nd March
💡 Got any ideas or feedback on how to improve this weekly digest? Just hit reply.
Acquisition
🐻 Satish Kumar, CEO. MS Eng Article
Closed Rounds
🧠 Imperative Care $100M 🇺🇸 Stroke-focused medical technology. 💰 Bain Capital Life Sciences, Catalio Capital Management, D1 Capital Partners
🧩 Doro Mind $4.7M 🇺🇸 Mental health caregiver platform. 💰 Berkeley SkyDeck Fund, City Light Capital, Precursor Ventures
🐻 Stephanie Greer, Co-Founder & Director of Product. PhD Neuroscience Article
🔐 Partisia Accelerator 🇩🇰 Privacy-preserving computation platform. 💰 Bitscale Capital, Insignius Capital
🐻 Kurt Nielsen, President & Co-Founder. PhD Economics
Date Built By Berkeley Started | Companies Funded | Total Raised ($M) |
7/8/24 | 691 | 209,981 |
Our goal is to document the startup ecosystem of Berkeley-founded companies. Please share this newsletter with any Cal Bears in your network so we can crowdsource information about all investment rounds and job opportunities.
Did we miss a company or want to announce a round? Add it here, and we will post it next week.
Do you have a job you want to post from a Berkeley Company? Add here.
Is there someone in the Berkeley ecosystem that you would want us to do a profile on? A founder, funder, or general startup person? Add here.
Built By Berkeley is not affiliated with UC Berkeley, but maybe we will be one day if we get enough subscribers….
Built By Berkeley, where we announce all the funding rounds by Berkeley-founded companies. This is a community effort, so please let us know if we missed a company here. 🐻
