ExpertsFMa16z's First Employee on the History of a16z, Venture Capital & Fixing the Government | Scott Kupor
David Senra · November 4, 2026
5 Selected Lessons:
framework · 3 case · 2
Disaggregating the super-connector role (CAA 'phalanx' model applied to VC)
Source
David Senra — a16z's First Employee on the History of a16z, Venture Capital & Fixing the Government | Scott Kupor (November 4, 2026) · at 3:26 ▸
When this applies
when a business's core value proposition depends on one irreplaceable, unscalable expert relationship-holder
The takeaway
When a critical function is concentrated in one irreplaceable person, disaggregate that function into specialized roles that can be staffed and scaled independently, surrounding the generalist with a team rather than depending on them alone.
Why it works
A single super-connector's capacity is bounded by their own hours and relationships; hiring specialized domain experts to cover each type of relationship (talent, BD, policy, etc.) around each generalist removes the bottleneck and lets expertise scale with the growing complexity of domains served.
One level up
Scale an organization by converting personal expertise into institutional systems before the expert becomes the bottleneck.
Where it stops holding
This stops working when the domain is narrow and stable enough that one generalist's relationships remain sufficient, making the added specialist layer unnecessary overhead.
Detail
procedure:
- Identify the single-point-of-failure expert whose personal network/relationships are the product's core value (e.g., John Doerr at Kleiner Perkins, or an agent at CAA).
- Break down everything that expert does into discrete domains (talent, BD, government relations, sales/marketing, etc.).
- Hire dedicated domain specialists for each discrete function.
- Let the original expert (now a GP) focus only on investing/judgment, surrounded by this team ('phalanx').
- Pitch the entire team to the client/entrepreneur as the value proposition ('reverse pitch'), not just the individual.
benchmarks: {"growth":"a16z grew from 3 people to ~100 in year one, to ~700 today"}
rationale: As the domains needing expertise multiply (software eating more industries), a single relationship-holder cannot keep up, so institutionalizing their function into specialized roles is the only way to scale.
In their words
So the fundamental theory that Mark and Ben had was, let's hire GPs who are functional domain experts in the particular areas of investing they want and then let's surround them with this failings of somebody who all they do all day long is manage CFO relationships or CEO relationships or VPs of engineering, somebody who all they do does government relations or business development or other stuff like that.
'Capital plus plus' — competing by changing the category instead of the existing rules
Source
David Senra — a16z's First Employee on the History of a16z, Venture Capital & Fixing the Government | Scott Kupor (November 4, 2026) · at 18:04 ▸
When this applies
when a new entrant lacks the track record/reputation incumbents compete on in any industry
The takeaway
If you cannot win on the current axis of competition, change the axis entirely and market loudly to establish the new category, since marketing itself is unnecessary only when scarcity already does the selling for you.
Why it works
Because the new entrant cannot win on the incumbents' existing axis of competition (historical track record), they instead redefine what is being sold (capital + services) and market aggressively to create a new category where their relative strengths matter.
One level up
Structural disadvantages are overcome not by imitation but by redefining the game.
Where it stops holding
This stops working when the new category promise is not actually delivered on, or when the market cannot distinguish genuine differentiation from marketing noise.
Detail
procedure:
- Recognize you cannot compete on the incumbents' currency (here: decades-long track record/brand).
- Identify what you can offer that incumbents structurally can't or won't (full services layer beyond capital).
- Build and deliver that offering (hire domain experts, build networks).
- Market it explicitly and loudly, breaking industry taboos against marketing if necessary, to create a new category label.
- Deliver on the promise to make the category credible.
rationale: Old-guard VCs didn't need to market because capital itself was scarce and did the selling; once capital became abundant, differentiation had to come from services, and that had to be actively communicated.
In their words
So what they rightly did was they said, look, if we can't compete on what people are competing on today, we have to change the rules of the road and go compete on something else. So they created this category was, okay, the new venture capital firm is going to be not just capital, but it's going be capital plus plus.
Decentralizing a scaled firm into domain-specific vertical teams
Source
David Senra — a16z's First Employee on the History of a16z, Venture Capital & Fixing the Government | Scott Kupor (November 4, 2026) · at 21:48 ▸
When this applies
when an organization's growing breadth of specialized needs outpaces what a single centralized structure can serve well
The takeaway
Decentralize operations into domain-specific units once the diversity of what you serve exceeds what one centralized structure can competently support, keeping centralized only the functions that are genuinely uniform across units.
Why it works
Centralization works while the domain is narrow and shared needs dominate; as the number of distinct domains grows, each with different support needs, a single centralized team cannot develop deep enough expertise in all of them, so splitting into domain-specific teams with their own tailored support restores service quality, while keeping only truly shared/compliance functions centralized.
One level up
Organizational structure must track the actual diversity of the problem space, not the org chart's history.
Where it stops holding
This stops working when the domains served remain similar enough that centralization still captures economies of scale, or when decentralization fragments relationships (e.g., LPs) that genuinely want a unified single point of contact.
Detail
procedure:
- Identify that one centralized structure can no longer serve the growing diversity of domains (software, crypto, bio, American dynamism, etc.).
- Create smaller, domain-specific investing teams led by domain experts (e.g., crypto fund, American dynamism fund, AI infra fund).
- Feather the ancillary operating/support services into each vertical so support matches each domain's actual needs.
- Keep only genuinely shared, compliance-sensitive, or relationship-continuity functions centralized (legal, finance, investor relations).
rationale: The needs of a gaming entrepreneur differ completely from a life-sciences entrepreneur, so one-size-fits-all operating support stops being useful past a certain scale/diversity.
In their words
we started a process ten years into the firm where we basically decentralized the firm. And we said, look, like we're going to have smaller investing teams that are going be domain specific.
Early fund wins (Skype, Nasira) compounding into fundraising and brand momentum
Source
David Senra — a16z's First Employee on the History of a16z, Venture Capital & Fixing the Government | Scott Kupor (November 4, 2026) · at 13:01 ▸ · at 15:53 ▸
When this applies
when an early, risky bet either compounds credibility or kills a nascent venture
The takeaway
In a new venture's earliest period, a rapid, visible success — even one that deviates from the stated strategy — can generate momentum that compounds disproportionately to its size.
Why it works
Fast, visible early wins create a news cycle and narrative momentum that lowers the cost of raising subsequent capital/customers, regardless of whether the win was strategically 'on thesis'.
One level up
Momentum begets momentum: the market rewards perceived trajectory as much as substance in the early innings.
Where it stops holding
This stops working when an early bet fails publicly instead of succeeding, which can just as easily destroy nascent credibility — the mechanism is symmetric risk, not guaranteed upside.
Detail
situation: a16z, just after closing its first fund and pitching itself as an early-stage investor, did an unconventional buyout of Skype from eBay alongside Silver Lake, working around an IP dispute with the original founders.
decision: They proceeded despite criticism that a buyout contradicted their stated early-stage strategy, and despite the broader VC industry hunkering down during the financial crisis.
outcome: Sold Skype to Microsoft 18 months later for 4x the purchase price, returning half the fund to LPs; combined with Nasira's quick sale to Cisco, this funded their services build-out and let them raise fund two ahead of schedule while competitors were paralyzed.
lesson: Being willing to act boldly and off-script when others are frozen (here, during a crisis) can generate outsized, narrative-shifting wins — but it is inherently a bet with no guarantee of this outcome.
In their words
And so we were able to actually like raise our second fund well ahead of what we had planned doing in part because the Skype thing just worked out so well. We returned, you know, literally, you know, half our fund to our LPs within eighteen months, which, you know, these days is obviously completely unheard of and even back then was unheard of.
'Ed' (Big Balls) rejected by rigid degree/experience proxy requirements
Source
David Senra — a16z's First Employee on the History of a16z, Venture Capital & Fixing the Government | Scott Kupor (November 4, 2026) · at 43:10 ▸
When this applies
when any hiring process uses credential proxies (degree, years of experience) instead of directly testing the actual skill needed
The takeaway
Replace credential proxies with direct skill assessment wherever possible, so hiring and leveling track actual demonstrated capability rather than indirect signals.
Why it works
Credential requirements are proxies for skill that correlate imperfectly with actual capability; when the proxy and the real skill diverge (a self-taught, highly capable 19-year-old developer with no degree), rigid proxy-based gating excludes genuinely high performers and underpays those it does let in.
One level up
Proxies are useful substitutes for unmeasurable things, but should be discarded the moment you can measure the real thing directly.
Where it stops holding
This stops working when direct skill assessment is infeasible or unreliable for a role, making credential proxies the best available signal despite their imperfection.
Detail
situation: A 19-year-old self-taught, highly skilled software developer ('Ed'/'Big Balls') joined DOGE/government work; a government employee publicly stated they would never have hired him because he lacked a college degree and five years of experience.
decision: Kupor used this as the rationale for OPM policy reform: eliminating blanket degree requirements, introducing third-party skills assessments, and allowing hiring at appropriate pay levels (e.g., GS-14/15) based on demonstrated merit rather than default entry-level placement.
outcome: Policy change in progress at OPM to let merit-tested hires enter at levels matching their actual skill rather than the lowest entry tier.
lesson: Rigid credential gates can exclude exactly the high-performing, unconventional talent an organization most needs.
In their words
He said, we would never hire Ed in government because he didn't have a college degree and he didn't have five years of work experience