Alpha in the age of agreement
Gil Dibner

Last week, over a stream of hipster coffees and expensive bottled waters, I had a series of catch-up meetings with VC colleagues and VC friends in the Bay Area. In every conversation, a now-familiar set of topics lines itself up for discussion: the rising valuations and sizes of Seed and Series A rounds (often starting at $100M pre-money), the frenetic and rapidly accelerating pace of investment decisions (measuring in hours), and - crucially - the narrowing set of signals that could conceivably assemble themselves into an investible thesis.

Anecdotes from a bubble. Across the four days, a few anecdotes stand out: stories of VCs staking out AI labs on campuses around the country; stories of deals closing in a matter of hours over a weekend with no product; stories of outstanding sales leaders refusing work at freshly minted unicorns, convinced that their exploding revenue is a leaky bucket of churn; and—most poignantly—the successful, exited, repeat founder telling me in a moment of honest clarity that “it’s just not any fun anymore” because his domain (like nearly every other) feels crowded with “100 people all running down the same path” as him.

Rationalization. When you talk to VCs these days - they use a few tropes to defend their decision-making. Sometimes, “revenue is just exploding” seems to be both true and compelling, but in too many cases there are nagging questions about durability, churn, and uniqueness. In a lot of cases, “the founders are just amazing” despite no revenue at all has become the fallback logic, taking the place of any robust analysis of product, technology, or competitive landscape. There isn’t really time for any of that. And finally - and I heard this from at least two experienced VCs - the rush to invest is justified by a rush to get in and out before the acknowledged bubble explodes in our face. One VC said five years, one said two, but they both want to play the game on the field and ride the wave before it’s too late. There will undoubtedly be some massive winners in the AI-fuelled race to reinvent software, and so this gold rush is not at all irrational.

The age of agreement. What is striking to me is the widespread wall-to-wall agreement of so many decision-makers in the tech business right now. There are a few truths that seem obviously true, and they are clear to just about everyone. Moreover, we have all realized them at more or less the same time.

• We all agree that AI will profoundly reshape business. It already is.

• We all agree that LLMs, multi-modal generative models, and reasoning models are powerful tools that unleash new value vectors.

• We all agree that prosumers and enterprises are eagerly lapping up these new products, resulting in unprecedented revenue growth for many players.

• We all agree that downstream investors (Series B and beyond) are willing to follow on with ever larger check sizes.

• We also all agree that founders that are (1) located in the right place (SF), (2) correctly pedigreed (the right lab or startup hyperscaler), and (3) part of the right networks (YC, Stanford, Neo, ZFellows, Prod).

There appears to be only one game in town right now and one way to play it: find a team of legible, pedigreed founders to build the next big AI application and focus on raising a big follow-on round almost immediately. Because everyone is playing the same game, they are chasing the same opportunities. Founders are going after the same spaces with similar products. VCs are going after the same narrow set of founders. All participants in this game understand the rules, and - thus - prices get bid up extraordinarily high and very quickly. Only time will tell if exit sizes will be large enough to justify the entry valuations and the shotgun wedding dynamics that are governing the landscape. As in every bubble, a handful of outlier cases will mask the realities of billions of incinerated capital. And as in every bubble, by the time the bubble dynamics are clear, the winners have typically long-since been minted. 

Alpha in the age of agreement. When the entire market agrees on nearly everything, it’s difficult, frustrating, and painful to be contrarian. To generate real alpha, one has to be short-term contrarian, long-term right, and - crucially - close enough on the timing that the contrarian can flip to consensus within a reasonable timeframe that works with fund lifecycles and cashflows. Doing this well is tricky at any time, but it’s particularly tricky in an age of agreement. The zeitgeist pushing everyone to focus on a few ideas is strong - and it’s not necessarily wrong. So follow-on investors and even customers might be more reluctant to bet on something that seems off-piste. Moreover, the reputational cost of being wrong is higher when everyone else is in violent agreement. Finally, it would be imprudent for any investor to build a portfolio comprised solely of one type of risk (either consensus or contrarian). 

At Angular, we’re trying to approach this unique time with enough mental flexibility to make smart investment decisions across the matrix of possibilities, which I’ve sketched below.

Historically, we’ve done our best and most rewarding work when we’ve invested in pre-traction and illegible companies. That said, we have invested in known-good (sometimes repeat) founder teams in the past. We’ve also increasingly focused on some traction at pre-seed, especially when a company is doing something exceptionally weird. The weirder a company is, the less likely it is to easily raise follow-on financing. In such cases, we believe that some early traction is both a signal of potential future follow-on financing and a cash-flow insurance policy in case that follow-on capital takes too much time to arrive. We’re trying to strike a balance. We think it would be foolish to slavishly avoid any investment that stumbles too close to consensus and starts to attract heat. At the same time, we wouldn’t be venture capitalists if we didn’t maniacally pursue the hunt for “true alpha” in its purest form. 

We are living through a very unique moment. This age of agreement will not last forever, and we are already seeing some early cracks in the wall of consensus - at least at the lowest technical levels. Wherever you are on this opportunity matrix, we’d love to talk. If you’re building at the center of the zone of agreement, we’d love to learn from what you are seeing. And if you are building something weird out on the edges, we’re super intrigued and eager to help. Either way, please reach out. It’s never too early, or too weird.

The AI-Native Enterprise Playbook
Ten real-time observations on a rapidly evolving playing field

No More Painting by Numbers
It’s the end of the “SaaS playbook.

The Age of Artisanal Software May Finally be Over
Every wave of technological innovation has been catalyzed by the cost of something expensive trending to zero. Now that’s happening to software.

Founders as Experiment Designers
David on why founders should run everything as an experiment.

ENTERPRISE/TECH NEWS

Wholesale on the rise. U.S. electricity prices have risen ~35% since 2020, and a deep dive into wholesale markets shows why: costs of generation (especially gas) spiked in 2021–22, and transmission bottlenecks are increasingly making cheap power “trapped.” Wholesale prices have nearly doubled in most regions, with congestion costs widening across ISOs like PJM and SPP, while California stands out as an exception with flat wholesale but still-rising retail rates. For startups and investors, this underscores that the real bottleneck in the energy transition isn’t generation cost curves - it’s the grid, suggestion transmission, storage, and demand-shaping infrastructure may be the next frontier for innovation.

Investor Nvidia. Nvidia is investing $5B into Intel to co-develop chips for data centers and PCs, marking a rare show of confidence in a struggling rival. The move follows $16B in lifelines from the U.S. government and SoftBank, suggesting Intel’s turnaround is becoming a coordinated geopolitical and industry priority. Nvidia also announced that they’re investing $100B in OpenAI. This investment is structured as a progressive, non-controlling stake alongside OpenAI’s purchase of millions of Nvidia processors. The use of the capital is to build at least 10 GW of AI data centers, with the first gigawatt slated for 2H 2026 on Nvidia’s Vera Rubin platform.

What can we learn from Chinese AI? An interview-based deep dive maps the “topology” of China’s AI ecosystem through a bilingual open-source engineer: beneath familiar model labs sits a vast, systematized infra workforce and a circle-driven collaboration model that’s structurally different from Silicon Valley’s “arrow-convergent” champion stack. The piece argues U.S. observers underrate China’s below-the-waterline capacity (fiber, data centers, cluster ops) and mislabel progress as “theft,” while China’s competitive edge comes from “cluster competition” (many strong contenders) plus policy-guided standard-setting—paired with a more pragmatic, less AGI-pilled culture. For founders and VCs, the signal is where moats may really form: software-hardware co-optimization, efficient cluster management, and partnerships across government/academia/big-tech circles, implications for GTM, vendor selection, and defensibility that go well beyond model benchmarks or agent hype.

HOW TO STARTUP

Probabilistic software. This piece is a month-old, but I’ve come back to it multiple times already. Gian Segator, an engineer at Anthropic, argues that we’ve crossed from deterministic software (funnels, predefined inputs) into a probabilistic era where models accept open-ended prompts and return distributions, making old PM/eng playbooks actively harmful. The prescription is to run AI companies like labs: optimize for “minimum viable intelligence,” embrace rewrites when models change, and build a unified data function that clusters real user trajectories (not just funnels) to manage cost, reliability and emergent behavior. For founders and VCs, the moat shifts to downstream data + empirical ops. Teams that think in probabilities, instrument outcomes, and restructure orgs around experimentation will compound; those forcing wave functions into classical dashboards will stall.

Virtual Agent Economies. A new paper from Google DeepMind proposes a framework for “sandbox economies” where autonomous AI agents transact and negotiate with one another at speeds far beyond human oversight. The authors argue that such virtual economies are already emerging and could either remain relatively impermeable testbeds or become deeply entangled with the human economy…raising risks of flash crashes, inequality, and regulatory arbitrage, but also opportunities for mission-oriented coordination (e.g. accelerating science, managing energy grids). For founders and investors, this frames a coming wave of infrastructure demand, not for chips but for, e.g., market mechanisms, currencies, and governance systems that will keep agent economies legible and steerable.

AI-designed life? Scientists have created the first viruses designed entirely by AI, synthesizing bacteriophages that successfully killed antibiotic-resistant E. coli strains. Researchers trained models on millions of viral genomes to generate and test new phage designs, with 16 showing targeted efficacy. For startups and VCs, this milestone signals the arrival of AI-driven synthetic biology platforms, tools that could reshape therapeutics, drug discovery, and even bio-manufacturing, while also raising profound biosafety and regulatory challenges.

HOW TO VENTURE

AI rollups are dumb. Dave Friedman argues that AI-enabled rollups are a mirage: the same structural reasons most PE rollups fail (overpaying, cultural friction, integration debt) still apply, and AI doesn’t dissolve them. While the VC pitch is that owning the “boring business” lets you plug in AI agents and expand margins, messy data, liability, and commoditization mean execution risk remains the core bottleneck. For founders and investors, the lesson is that real rollup success comes from decades-long discipline and permanent capital, not the VC “growth at all costs” mindset, making AI rollups a mismatch for venture timelines.

Firebold appoints Hemanth Vedagarbha to CEO as Global expansion accelerates.

Groundcover launches zero-instrumentation for LLM observability.

Valohai MLOps platform now available on Oracle Cloud Marketplace.

Aquant
Digital Marketing Manager (NYC, hybrid)

Candu
Full Stack Developer (London, remote)

Motorica
Platform Engineer (Stockholm)