Partnering with founders for the hardest things
Gil Dibner

Over the past week, I did my best to help one of our portfolio founders through an extremely difficult situation. The details really don’t matter, but the severity was extreme on both the emotional and business levels. As a business matter, the situation threatened to destroy years of work and quite a bit of value. It threatened to rip apart a small team that had been assembled over years - and on some level it did that. On an emotional level, I can’t express how challenging the situation was for everyone involved. Fortunately, we had the right experienced people around the table, and it looks like we’ll get through it. But the past few weeks and the coming few weeks have been and will continue to be harrowing.

On the basis of cold numbers, this is probably not where I should be spending my time. The company involved is not - yet - a success story. Like most venture stories, it’s unlikely to be a huge win, and, given recent challenges, the odds of success are longer right now than for many other startups. But for the founder(s) involved, this is a big deal. It matters. There are customers, jobs, and relationships on the line. There is the emotional stress of seeing years of work potentially go up in smoke. There are legal questions, there are business questions, there are management questions, there are moral and ethical questions, and - in short - it’s just really really hard. It’s more than hard. It’s emotionally painful.

Entrepreneurship (and venture) is often seen as a hits-driven business, which - of course - it is. We all know about the “power law” and the idea that one huge outcome is supposed to more than compensate for the risk across a venture portfolio. The wise advice from my more experienced venture colleagues is to learn to write companies off and to do it as early as possible. They have a point. We are in the business of driving LP returns, and not rescuing flailing businesses by pouring good money after bad. Precious time should be deployed just as carefully.

But in practice, it’s rarely as neat as the theory suggests. There are real people tied to each company in a portfolio, including the ones that are struggling. Those people have relationships and aspirations. There are intangibles such as grit and determination that are really hard to fit into a venture portfolio model. In the immortal words of Miracle Max (Billy Crystal) from The Princess Bride, “there is a big difference between mostly dead and all dead. Mostly dead is slightly alive.” To be clear - VCs are not miracle workers. Unlike Miracle Max, we can’t raise a startup from the dead. But some founders can, and doing so is painful and very very hard.

Some VC websites don’t refer to when they invested in a company. They refer to when they “partnered” with that company instead. While it’s a bit of a tired cliche, I think it still captures some of the essence of what true venture capital is (or should be) about. Early stage VCs really should view themselves as true partners of founders. That means that we must be available to participate in the hardest moments and most difficult discussions and decisions - even when the odds of success are vanishingly small. These very hard moments are not why any of us (investors or founders) got into the business, but they are an integral part of serving as a true partner to founders. They also happen to be a part of many entrepreneurial success stories as well. There are certainly many cases where the cold calculus prevails and where a shut down to avoid a fruitless struggle is the wisest course, but there are other cases that demand and deserve continued partnership and investment of time (and sometimes capital) to help the right team of founders battle through against truly impossible odds. Perhaps this willingness to remain partnered through such challenging and painful situations is an integral part of the optimism so many VCs talk about as essential to their worldview and - ultimately - their success.

No Sleepwalk to Success
Engineering success in a technical startup.

Revenue Durability in the LLM World
Everything about LLMs seems to make revenue durability more challenging than ever.

A Robotic Future for Retail Grocery, Finally.
Why we invested in Finally.

A Digital Fabric for Maritime Trade
Why we invested in Portchain.

More blogs

UK / FinTech. CloudPay raised $120M, led by Blue Own Capital, to accelerate the recent growth of their long-standing payroll processing platform.

UK / Semiconductors. Riverlane raised $75M, led by Planet First Partners, for its “quantum error correction” technology platform, after receiving a surge of interest from quantum computing customers.

Germany / SaaS. KNIME raised $30M, led by Invus, for its no-code data analytics platform.

UK / SaaS. Datch raised $15M, led by Third Prime, for its AI-powered “frontline intelligence” platform, targeting heavy industry.

Netherlands / Nanotechnology. VSParticle raised $6.5M, led by NordicNinja and Plural, for its nanoparticle creation process which enables clients to produce custom nanoparticles on demand.

ENTERPRISE/TECH NEWS

Monopoly money. As reported in the Wall Street Journal, “a federal judge ruled that Google engaged in illegal practices to preserve its search engine monopoly,” granting a major antitrust victory to the DOJ. The central argument of the DOJ is as follows: Google suppressed competition for its search engine by paying operators of web browsers and phone manufacturers to make Google the default search engine. In other words, Google may have won its monopoly through innovation, but it can’t buy its way into a continued monopoly. So the DOJ won. But what can be done? That’s the central question, as articulated by Ben Thompson over at Stratechery. Thompson argues that Google ought to be disallowed from making any sort of payments or revenue share for search, which will (hopefully) have the result of making the space more competitive (and better for consumers), just like what happened when Google and Apple both launched Maps products.

AI without OpenAI. After the failed leadership coup at OpenAI, Microsoft has worked to execute an AI strategy independent of OpenAI and all its associated drama, so reports the Financial Times. This is a must-read deep dive on Microsoft’s reaction to OpenAI’s leadership crisis, and what they’ve done to shore up their strategy since then. Notably, they were the first player to execute on the non-acquisition acquisition when they convinced Inflection CEO Mustafa Suleyman to leave and join Microsoft (like Google’s deal with Character.ai and Noam Shazeer a few weeks ago).

HOW TO STARTUP

Glean on the rise. Is Glean one of the most interesting enterprise AI stories out there right now? I think so. Glean, the enterprise search product company, is raising now at double the valuation of just 6 months ago. Glean is a great example of a company that has layered AI onto its initial wedge (enterprise search) to create unique and compelling AI-powered product experiences. Many other companies have tried to create assistants or AI-powered app creation platform, but Glean has been able to move so quickly because they were already sitting on top of their customers’ enterprise data.

AI advancements continue. A few interesting data points and anecdotes on how quickly AI is advancing as of late. Ahmed Awais, CEO Langbase, shares that 34% of their “pipes” have moved from OpenAI to Sonnet or Llama 405B…reinforcing the idea that these models are quickly commoditizing and customers will find their way to the best performing / cheapest model. Paul Graham shares another anecdote from a friend who estimated that price per unit of performance has decreased about 100x in each of the past two years. As Graham writes, “that’s a level of change that makes the future extremely hard to predict.”

Future of software engineering. Worth reading this thread on the future of software engineering from Russell Kaplan, president of Cognition Labs (makers of Devin), in full. In particular, I wanted to highlight Kaplan’s prediction that, as agents do more of the coding, then companies will start marketing to agents. Kaplan also argues that half-baked or feature-incomplete MVPs will be less acceptable (as we’ve argued here) and that testing infrastructure will have to be much more robust. Again, the whole thread is worth a read!

HOW TO VENTURE

OpenAI exodus. 10 of the original 12 members of the OpenAI team have now left. I wonder what the returns would be for a fund that could back them all?

Model mania? A great rundown from Eric Newcomer on what’s next, according to top investors, after foundational model mania recedes. A few key takeaways (not surprising to readers of this newsletter, but useful to highlight nonetheless). First, VCs are no longer interested in companies building their own foundation models, especially given Meta’s willingness to pour billions into open source. Second, lots of infrastructure, but no breakout applications, is getting everyone nervous about making bets at the model or infrastructure level. AI-powered software, where AI is just one ingredient, is more of a focus.

LightSolver has been awarded €12.5 million in funding as part of the European Innovation Council Accelerator Program.

Groundcover introduced OnPrem and airGapped, a modern observability backend tailor-made for offline environments.

Paradime is speeding up data pipeline for customers up to 50% without increasing costs using AWS Graviton.

Lunar.dev is hosting a happy hour on September 16 in San Francisco on "The Future of AI and APIs". Apply to attend here.

CruxOCM
Engineering Manager (Remote)

Firebolt
Product Manager (Remote)

Planable
QA Automation Engineer (Moldova)

Vault Platform
Full Stack Software Engineer (UK Remote)

More jobs