The Atoms Gap

Between August and November 2026, 841 venture capital blogs and newsletters published 1,978 posts. a16z, Sequoia, USV, Lux, solo GPs, operator newsletters. I read all of them, side by side. 586 were news, fund announcements, team pages or event invites. That left 1,392 real ideas from 199 sources.

I wanted to know what the people who fund the future actually think about right now. Not what one of them says on a podcast. What all of them write, when you put it next to each other.

Every idea got two scores. The venture score asks what a good investor asks: is the opportunity specific, is the market big, is there a reason it works now, is there a moat, is there evidence, does it go against the crowd, can a founder act on it. The techno-optimism score asks something else: if this works, how much does it change the world. Ambition, whether something important gets much cheaper, how much new technology it needs, whether it touches the physical world, and whether the author argues for building or for slowing down.

Venture mostly writes about itself

The biggest category is venture capital itself. 225 of the 1,392 ideas, 16 percent, are about funds, LPs, seed prices, valuations, exits, how to be a better VC. Add markets and macro, interest rates, the AI bubble debate, and you are at 26 percent. One in four ideas from people whose job is to fund the future is about the money, not about the future.

AI and software take a third, mostly the economics of software. Who captures the value, seats or outcomes, what happens to SaaS.

And then there is the part that I care about. Energy, robotics, biology, industry and defense, mobility. Everything that is made of atoms. Together that is 123 ideas, 8.8 percent.

The atoms gap

Here is the thing that I did not expect. When you plot attention against the techno-optimism score, the categories sort themselves.

Venture itself and markets sit on the far right, lots of attention, and at the bottom, techno-optimism 0.21 and 0.16. Robotics is the highest of all twenty categories, 0.53. Bio and health 0.49, industry, defense and space 0.48, energy 0.42. That's the gap.

You could say of course, a post about seed valuations is not supposed to be ambitious. That's fair. So look at the venture score, which does not reward ambition at all. Atoms ideas score 0.57 there, venture and markets 0.46. The atoms posts are not only more ambitious, they are also more specific and better argued, by the investors' own standard.

Industry, defense and space has 18 ideas in four months. It scores 0.63 on the venture score, the third highest of any category. Eighteen.

It has been like this for ten years

Four months could be a fluke, so I went back further, to every post the same blogs and newsletters published since 2016, 55,938 of them.

Venture and markets have been between 26 and 34 percent of all posts in every year since 2017. AI went from 9 percent in 2020 to 29 percent in 2026. Atoms peaked at 12 percent in 2021 and 2022, in the climate boom, and fell back to 10 percent. So the gap is not new. Venture has always written mostly about venture, and atoms never got more than about one post in eight.

Everyone is looking at now

The second pattern is time. For every idea: if the main claim comes true, when does most of the impact arrive.

56 percent of ideas are about now, the next twelve months. Only 37 ideas, 2.7 percent, look more than ten years ahead. For a business that sells ten-year funds, that is a short view.

The atoms ideas are the exception again. 59 percent of them look three years or more ahead. Across all ideas it is 22 percent.

Where the herd is

Some ideas show up again and again. I grouped the posts that make the same specific claim, not only the same topic.

The strongest consensus by far: models are becoming a commodity and the value moves to the application layer. 14 different sources argued this between August and November. a16z was first, on August 10.

Second, with 7 sources, AI companies should sell the work, not the software, priced on outcomes instead of seats. Third, with 6, venture is crowding: too much money in the same few companies, too much consensus, valuations that do not match reality.

So six different venture writers say venture has a consensus problem, in the same three months. I think that is a little funny.

Fifth is the one closest to my own view. Software got cheap, and venture returns to atoms and science. Five sources. Jared Heyman's analysis of the YC summer batch, Alex Oppenheimer, Akshay Mehra, Obvious Ventures, Alumni Ventures. So the shift is being talked about. It is just not yet being written about very much.

Only 182 of the 1,392 ideas, 13 percent, belong to a thesis that two or more sources share. Most ideas stand alone.

The contrarian shelf

That leaves the other side. Ideas that score high, that go against the common view, and that nobody else made in these four months. If you invest, this is where I would start reading. If you build, this is where nobody is crowding you yet.

1. Unburdened By What Has Been Andrew Ziperski · Venture itselfVenture capital is hitting its limits; the next generation of tech companies will require massive credit markets and asset-backed financing that megafunds must master to capture the coming $500B+ opportunity, not chase venture primacy.
2. Population Decline Is Real. So Is the Reversal. Fabrice Grinda · Society & policyPopulation decline appears inevitable but reverses through selection effects, children becoming luxury goods among elites, falling childcare costs, and AI/robotics attacking parental time constraints—though demographic momentum means fiscal pressure remains committed for 30 years regardless.
3. Defying Gravity: The System of Record in the Age of AI Battery Ventures · AI & software economicsAI is forcing enterprises to reconsider and replace legacy systems of record because they can't deploy AI agents without clean, consolidated data—creating a rare window for startups to displace deeply entrenched software incumbents across both horizontal and vertical markets.
4. Anthropic Isn't Just a Company, It's a Payoff BIG by Matt Stoller · Markets & macroStoller argues Anthropic's impending IPO will primarily benefit Google and Amazon despite their competitive positioning, suggesting the company masks fragmentation across multiple failing monopolies rather than genuine AI innovation.
5. AI, Owning the OpCo, Who will capture the value? Erik de Stefanis · AI & software economicsOperating companies that own and deploy AI-enhanced custom software will capture more durable value than software vendors, whose moats are eroding; the key is rebuilding software around actual workflows, charging for value delivered, and leveraging margin compression to create enterprise value that accrues to business owners rather than SaaS vendors.
6. Why Your Voice AI Agent Isn’t Enough Euclid Ventures · AI & software economicsVoice AI agents alone won't drive defensible value; vertical platforms must own the full telephony stack—line, agent, and record system—to capture call data and build sustainable competitive moats against carriers and incumbent systems of record.
7. The Rebuild, Part II: Outside the Walls Carried Away · Venture itselfVenture firms should redesign themselves from first principles around AI, treating a fully automated operation as the baseline and adding humans back only where machines demonstrably fail—rather than bolting AI onto existing structures. Incumbents can spawn new entities with fresh org designs while preserving brand and relationships.
8. The Durable Majority Jess Schram · Venture itselfAI is enabling a new class of profitable, capital-efficient software companies that venture is ill-equipped to fund. The author proposes a hybrid structure—priced equity plus revenue-share repayment—to capture durable mid-market exits while delivering faster, more predictable LP returns than traditional venture.

The best ideas on each score

Highest on techno-optimism:

1. Data At The Edge USV · Data & knowledge
2. Airbound: As We May Move Not Boring · Mobility & transport
3. The Cognitive Revolution Sequoia · Society & policy
4. When Physical Work Becomes Programmable Bilal Zuberi · Robotics & physical AI
5. Population Decline Is Real. So Is the Reversal. Fabrice Grinda · Society & policy
6. From Software to Science: Venture Capital’s New Frontier is Atoms Obvious · Venture itself
7. Investing in Form Energy: America’s Battery Champion for the AI Grid David Cahn · Energy & climate
8. How Did We Get Here? Akshay Mehra · Venture itself

Highest on the venture score:

1. Getting to General Robots That Don’t Suck Annelies Gamble · Robotics & physical AI
2. Vertical B2B Leader Owner Acclerated After $100M ARR With Agents. The Key Insight: Every Time a Customer Logs In, Owner Has Failed. Saastr · AI & software economics
3. Compute Capital Markets Multicoin Capital · Fintech & crypto
4. The New “Ask Your Developer” Is Ask Your Agent. And It’s The Marketing Battle You Have to Win Now Saastr · Sales & growth
5. AI, Owning the OpCo, Who will capture the value? Erik de Stefanis · AI & software economics
6. The Rise Of The Neo Neo Lab Aspiring for Intelligence · Vertical AI services
7. Why Your Voice AI Agent Isn’t Enough Euclid Ventures · AI & software economics
8. Defense in Depth, Rebuilt for Machine Speed A16Z · Security

Explore it yourself

Every dot is one idea. Right is the venture score, up is techno-optimism. The solid dots are atoms. Click a dot to open the original.

What I take from it

If you build in energy, robotics, biology or industry, you are working on the ideas that score highest on both counts, and you are competing with very few other voices for attention. That's a good position. The people who decide where capital goes are still mostly reading about each other.

If you invest, the herd is easy to see. Fourteen sources on the same thesis in three months is not an edge anymore. The edge is in the eight percent.

I wrote before that some problems need atoms, not bits, and that cheap energy changes everything. Now I can show it with what investors write themselves.

What this does not show

• Writing is not money. This is what investors write, not what they fund. A firm can write about seed prices and invest in fusion.
• Four months is a short window. The ten-year trend helps, but years before 2016 have too few posts to show.
• Sources are blogs, not firms. A partner's personal newsletter and the firm's blog count as two sources.
• Scores are judgments, not facts. A 0.53 is a position on a scale, not a law of nature.

Every idea above links to the original. The writing belongs to each author.