- The Angle
- Posts
- Dominant Narrative vs. Specific Reality
Dominant Narrative vs. Specific Reality
Gil Dibner
Last week, David wrote a brilliant piece that critically re-examined the idea that AI allows startups to “sell the work.” While potentially true, there are some critical challenges buried inside that business model. As David wrote, “what looks like a new equilibrium may simply be a momentary pause as business models evolve,” and - in the case of selling the work - the power of AI may be more deflationary than many of us expect. As David argues, startups that “sell the work” may end up driving the value of that work to zero faster than they are able to capture much of that value.
The dominant narrative. We are living through a unique time in the tech landscape. A single dominant narrative looms over the landscape. Any startup that can credibly claim to be riding the AI wave (AI-first applications) or driving the AI wave (foundational models or other AI infrastructure) seems to be able to attract tremendous amounts of venture funding at eye-watering valuations. That excitement has rewritten the dominant narrative of venture capital itself: AI is enormously valuable, therefore AI exits will be enormous, therefore huge seed rounds at high valuations for companies with no revenues can make sense, and therefore VC funds should be comfortable investing at high prices for low ownerships, and therefore VC funds should continue to get ever larger so that they can be players in the new VC game of AI. This, in turn, feeds back to founders who realize that if they only weave AI into their stories sufficiently, they can justify nearly any valuation. And maybe with a bit of secondary sales they can buy a house too.
The crowding out effect. The problem with such an overwhelmingly dominant narrative is that it appears to be crowding out any other narrative, however valid. Companies that are not (strictly speaking) AI-first are finding that unless they weave AI into their narrative, they can’t even get VC meetings. Sometimes (maybe even most of the time) this nudges them in the right direction. There are some specific ways in which AI is likely to impact nearly every software business, and ignoring that would be foolish. But there are still plenty of businesses where AI is either completely besides the point or where it will be easily leveraged without much effort. For those businesses, the core narrative that drives the business is simply orthogonal (not contradictory!) to the dominant narrative of the day. Because the entire ecosystem is driven by one dominant narrative, there is little room for other narratives on how value might be created.
Value vector uncertainty. As David pointed out in his piece (and as Aileen Lee compellingly argued recently as well), we are still in the very early days of figuring out how AI is actually going to create significant value delivery vectors for incumbents and new entrants. My guess is that in many cases, nimble incumbents will benefit more quickly, more easily, and perhaps more naturally than a lot of new entrants. Many of the benefits of AI lend themselves to incumbents: cheaper code for faster iteration, easier natural interfaces, easier generative outputs on existing data, and easier addition of intelligent features on top of existing workflows. (A recent podcast from No Priors with the CEO of Rippling provides a great example of this. For Rippling, AI-based analysis of employee productivity - and, hence, fire-ability - is just one more feature amongst a big set of compounding features.) Start-ups may face some interesting opportunities to explore and establish novel value vectors, but it’s not that easy to identify what those might be. Personally, I have a hard time convincing myself that the umpteenth vertical foundational model for legal document analysis or yet another model that can synthesize natural-sounding text-to-speech is going to form a sustainable barrier to entry. The dominant narrative, however, is so powerful that these companies are getting funded and scaling regardless. I am sure some of them will create sustainable value - but right now it’s hard to see how and it’s nearly impossible to predict winners. This uncertainty is unnerving. It’s easier to focus on the dominant narrative - the idea that there is a “well understood” path to a $10B outcome. That, however, is self-deception. There is more uncertainty about value vectors than ever before. This uncertainty should be music to the ears of both founders and VCs because it means the world has - once again - been broken wide open by a new technology and no one knows the answer. At a time of such great opportunity, the presence of an overwhelmingly dominant narrative is particularly unfortunate. We need creative independent thinking more than ever.
Ride the dominant narrative or forge a specific reality. Founders face a fairly binary choice: They can ride the dominant narrative or they can work to forge a specific reality that relates to their unique insight on their specific market. There are countless teams today that are building an “autonomous AI agent for X” or a “foundational model for Y.” They all appear to be hoping to ride the financing wave, but very few of them are armed with a unique insight or capability. Some will get funded, of course, and some subset of those might deliver sustainable value to customers. Time will tell. What is already clear is that the dominant narrative(s) around AI are so well understood and so well funded right now that it’s nearly impossible for an early-stage fund to play that game. While that might sound like bad news, I disagree. Our role in the ecosystem is to partner with founders that are driven by a unique non-consensus insight that happens to be right. Wall-to-wall consensus is the enemy of venture returns, but it can also have a dampening effect on authentic entrepreneurship. We are looking for those few founders who are able to combine some unique insight about a market with an idiosyncratic application of AI to build something non-consensus. Given how obsessed everyone is with a pretty narrow set of ideas, my suspicion is that the most interesting companies getting started right now are doing so with a great deal of uncertainty as they find themselves questioning those narratives. Those are the companies I’d most like to meet.
FROM THE BLOG
Am I Thinking About AI the Right Way?
Gil shares the four AI themes and questions he's thinking about.
The Venture Apocalypse
The venture world is deeply debating its future, but core principles remain unchanged.
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.
WORTH READING
ENTERPRISE/TECH NEWS
6.6B. OpenAI raised $6.6B, the largest venture round ever, last week. (And some argue they’ll need to raise another $50-100B next year). The round was led by Thrive Capital at a $157B valuation. The investment is contingent on OpenAI restructuring from a non-profit to a for-profit company. It seems that investors in this round were also asked to not invest in any of the foundational model competitor companies, including former OpenAI cofounder Ilya Sutskever‘s company SSI.
Meta is on a roll. Last week, Meta announced MovieGen, their answer to OpenAI’s Sora model and other players like RunwayML. It’s not ready for public use just yet, but the demos are incredibly impressive. Meta also released the latest Llama models - Llama 3.2 and 1B/3B sizes. Llama 3.2 is Meta’s first multimodal model, which is to say it understands both images and text. The 1B and 3B Llama models are super-small, optimized to run locally on devices (like a smartphone, or maybe someday soon, Meta’s Ray-Ban glasses).
AI is like tuition. An interesting take on AI from the CIO of CommonSpirit, a health system in the USA with 142 hospitals in 20 states and nearly 175,000 employees, Daniel Barchi: “Cybersecurity is like rent — you've got to pay it. There’s no choice; otherwise, you don’t have shelter. AI is like tuition — we're learning, there are benefits, but you’re investing for the future.” Important to recognize how far away we are from the promise of AI in many enterprise contexts.
HOW TO STARTUP
Leading lessons from Hubspot CEO. Brian Halligan, Hubspot’s cofounder and longtime CEO, has been posting a series of great threads on X about lessons learned from his journey with Hubspot. The latest, on leading, is a must-read.
Advisory equity benchmarks. Peter Walker from Carta shared some benchmarks on advisor equity this past week. You can find them here. Advisor equity is often hard to negotiate, and I find the FAST Agreement from the Founders Institute to be a bit too subjective to be useful. This, from Carta, is fantastic.
Marketing lessons from Mutiny. Ryan Narod, former head of Marketing of Mutiny, shared his top five lessons learned from his 3.5 year long journey leading the marketing function at Mutiny. Marketing is, in my opinion, an incredibly challenging function to get right. So Ryan’s lessons are worth a read. I especially appreciated his advice that all programs should have “distribution built-in.” (Far too often, marketing departments create great programs that nobody ever sees.)
HOW TO VENTURE
CRV giving it back. In what is immediately being heralded as a “sign of the times,” CRV announced that it will be returning $275M of its $500M Select fund (a growth stage-focused vehicle) back to investors. “The math no longer works” on the investments, given frothy valuations and a non-existent exit market. CRV also announced that it does not plan to raise another Select fund.
How does the math work? 10 high-flying foundational model companies, valued at a total of $21.75B, have a combined revenue of less than $100M. Some of these companies have great products, most of them have amazing talent. But, as Deedy from Menlo asks, how many will be great businesses? And, perhaps more to the point, how many will provide great returns for venture investors? It’s hard to see how the math works.
PORTFOLIO NEWS
Reco’s CPO and co-founder Gal Nakash shared three critical components that business leaders can use to keep their SaaS stack secure.
Forter published the second annual Consumer Trust Premium Report, which highlighted a strong connection between trust and the amount that consumers spend.
PORTFOLIO JOBS
Firebolt
Solution Architect (Munich)
Reply