Uncapped with Jack AltmanFrom BoxGroup to Benchmark and Back | Greg Rosen, Partner at BoxGroup | Ep. 22
CHAPTERS
- 0:00 – 0:47
Greg Rosen’s collaborative venture philosophy and why it’s “anti-scale”
Jack frames the conversation around BoxGroup’s “collaborative venture” approach and why it has become rarer as funds got larger and more competitive. Greg explains that true collaboration requires structural choices that limit certain kinds of scaling.
- •Collaborative venture works best when a firm can remain neutral and broadly helpful
- •The model is inherently “anti-scale” in some dimensions
- •Industry shift: bigger funds, more competition, sharper elbows, less syndicate-style collaboration
- •BoxGroup’s growth despite the broader decline of collaboration
- 0:47 – 5:47
Staying “Switzerland”: why BoxGroup won’t lead rounds
Greg describes the key design constraint: BoxGroup avoids leading seed/Series A rounds to preserve trust with other investors and founders. He explains how even leading a few rounds creates doubts about whether the best deals are being shared.
- •Leading even 1–3 rounds a year breaks the perception of neutrality
- •Other investors would suspect deal-sharing is incomplete or biased
- •BoxGroup scales check size with round inflation but avoids lead-check behavior
- •Goal: be the easiest “yes” in many different syndicate configurations
- 5:47 – 11:59
Scaling the model: AUM vs velocity (and why velocity is brutally hard)
Greg contrasts two ways to scale a venture firm: raise more capital and move upmarket, or do more early deals. BoxGroup chooses velocity—seeing and doing far more deals—despite the operational intensity this demands.
- •Two scaling paths: larger funds/leading rounds vs more deals at the same stage
- •BoxGroup scales by volume: ~70–80 investments/year
- •The workload is extreme; calendars at BoxGroup look very different from typical partners
- •Most firms choose the AUM path because the velocity path is operationally demanding
- 11:59 – 14:44
Solving adverse selection: trading ownership for access to the best companies
Jack and Greg dig into the core problem BoxGroup is optimizing for: avoiding adverse selection. Greg argues it’s better to own less of the right companies than more of the wrong ones, and that collaboration improves access quality.
- •Primary objective: avoid adverse selection more than maximize ownership
- •Tradeoff: lower ownership percentage in exchange for better deal flow and entry
- •Critique they hear: “you need big ownership” vs their view: “you need the right companies”
- •BoxGroup aims to get into more of the eventual category-definers and follow-on support them
- 14:44 – 21:10
“Seeing” as the main lever: humility about picking in a hyper-competitive market
Greg argues venture has become more competitive, and many investors respond by claiming they’ll become better pickers. BoxGroup instead assumes picking improvements are limited and focuses on the “seeing” variable—finding more great opportunities earlier.
- •Venture is harder: faster information spread, fewer durable proprietary deals
- •If competition increases and you don’t change your process, returns should fall
- •BoxGroup’s “venture humility”: don’t assume you can 2–10x picking skill
- •Operating stats: ~5,000–6,000 qualified opportunities/year → ~70–80 investments
- 21:10 – 23:25
How BoxGroup sees so many companies: outbound-first, high-quality relationship engines
Greg breaks sourcing into inbound and outbound, emphasizing that outbound is the “cake” and inbound is “icing.” He outlines BoxGroup’s event-driven and relationship-driven tactics, plus why quality and curation matter more than raw activity.
- •Inbound is helpful but insufficient; outbound is the core engine
- •Tactics: high-effort events (campuses, engineers, leaders), curated for quality
- •~25% of deals come from Switzerland-style sharing with other investors
- •Trust-based sharing works because BoxGroup won’t snipe leads and can recommend future leads
- 23:25 – 27:57
Getting to founders before the deck: investing in people before companies exist
Greg explains BoxGroup’s pre-seed edge: meeting future founders before they leave their jobs, often before there’s a coherent idea or pitch deck. This requires time-intensive meetings and comfort with ambiguity—and many meetings that won’t convert.
- •Internal mantra: “If there’s a deck, it’s too late”
- •Core activity: meet individuals pre-company, even pre-departure from their job
- •Accept many meetings that don’t turn into companies (directors who stay directors, etc.)
- •Early-stage ambiguity (lots of ideas, lack of cohesion) is often a feature, not a bug
- 27:57 – 31:46
Decision-making that fits pre-seed: no partner meetings, single-trigger, and “help to yes” culture
Greg argues traditional partner meetings and consensus voting are mismatched with seed-stage ambiguity and time constraints. BoxGroup uses single-trigger decisions in small pods and a culture that helps the deal sponsor articulate the bullish case rather than default to hole-poking.
- •Partner meetings are inefficient and performative for founders
- •Three buckets: obviously great, obviously bad, and the “middle” where alpha lives
- •Most big outcomes start in the “middle” bucket, not the obvious winners
- •Team norm: pull threads to help someone “get to yes,” not to veto on margin
- 31:46 – 34:39
What’s learnable about picking: taste vs reps, plus simple heuristics
Greg splits investing judgment into innate “taste” and learned pattern recognition through reps. He shares specific heuristics BoxGroup has developed by reviewing misses and observing their own decision dynamics.
- •Some “deal taste” is hard to teach; preferences and talent clustering matter
- •Reps are essential: seeing thousands of companies builds pattern recognition
- •Heuristic: if you take a third meeting, you should usually just do the deal
- •Other learnings: always meet the CEO; don’t over-index on early market precision if the person is exceptional
- 34:39 – 37:25
Calendar auditing: eliminating empty calories and spending time where alpha is
Greg and Jack discuss how venture’s ambiguity enables poor time allocation. Greg critiques common time sinks—especially VC-to-VC networking—and pushes for time spent either helping portfolio companies or meeting new founders.
- •VC networking events are framed as the biggest time-waster
- •People hide behind board obligations while still spending time on low-output activities
- •Discipline: say no often; protect time for founder meetings and real sourcing
- •The right calendar depends on strategy, but most people aren’t honest about outputs
- 37:25 – 41:03
Find your outlier advantage: the three investor archetypes and focusing your energy
Greg describes three archetypes—hyper-networker, philosopher/media, and specialist—and warns against copying others’ playbooks. The key is identifying where you’re an outlier and aligning tactics to that advantage, then auditing time accordingly.
- •Archetypes: hyper-networker, philosopher/media operator, and specialist
- •Copying someone else’s strengths is alluring but usually underestimates the difficulty
- •You can minor in other archetypes, but must major in your true advantage
- •Use calendar/output audits to confirm what actually produces results
- 41:03 – 44:00
Networks and trust cycles: from old VC cliques to modern “preferred partners”
Greg reflects on how power shifted toward founders, then how the market became noisy as capital and new firms proliferated. He suggests the ecosystem is evolving toward a new form of “preferred partners,” driven by founder choice and trusted recommendations.
- •Past: VC cliques and more investor control over next-round outcomes
- •Middle period: founder power increased, but the market got messy/noisy
- •Now: founders still have power, but rely on trusted investors to navigate options
- •BoxGroup plays favorites in recommendations based on observed partner behavior across its portfolio
- 44:00 – 45:28
Future bets: AI-native coding tools and the long arc toward brain–computer interfaces
In closing, Jack asks Greg for forward-looking views on code generation and brain-computer tech. Greg compares AI’s current phase to early mobile—initially porting old paradigms before discovering truly native primitives—and speculates about matrix-like input/output as a path of least resistance.
- •AI coding is evolving quickly; nobody fully knows the end-state
- •Analogy: early mobile copied web; later produced native breakthroughs (Uber/Instagram, etc.)
- •Cursor succeeds by upgrading the familiar IDE; Warp explores agentic development environments
- •Speculation: brain-computer interfaces might be a more direct route to “new worlds” than robotics/terraforming
