← All firm reviews
Refinery · Private research

Craft Ventures: strategic review

23 September 2026 · Deep public-evidence review · Laurie review pending

Craft Ventures: the operator claim is real, while the firm-level idea is still catching up

Craft Ventures is trying to own a demanding position: the B2B software investor whose former operators can transfer winning company-building practice into the portfolio. The evidence supports more of that claim than the homepage reveals. Craft has embedded operators inside Replit, Supabase and Horizon3.ai; specialist executive councils; concrete growth, talent and cyber playbooks; and founder testimony describing customer, recruiting and GTM help. Its strongest current work shows the labour behind the promise. Its public system is only partly effective for the job because that proof sits beneath a generic institutional frame and an article feed dominated by investment announcements.

The firm sits between enterprise-software specialists and platform-heavy multi-stage firms. It invests from early stage through growth, with a current concentration in AI applications, infrastructure and cybersecurity. Its distinctive territory is operator deployment around B2B growth. First Round owns reusable operator knowledge more strongly as an editorial destination. Bessemer owns cloud and AI benchmarks more strongly as a repeatable research franchise. a16z owns cross-format media scale. Craft can make a sharper claim than each of them when an operator has done the work inside a company and can expose the method with numbers.

What Craft appears to want to own

Three associations recur across the current institution and its history:

  1. B2B operator-investors who deploy repeatable scaling playbooks. This is the central and most credible present association. The team, OIR assignments, councils, founder stories and case studies all reinforce it.
  2. Disciplined SaaS and product-led growth economics. Burn Multiple, the SaaS metrics framework and SaaSGrid gave Craft a historically portable idea. Independent practitioners still cite and adapt Burn Multiple.
  3. AI-era cybersecurity, infrastructure and GTM expertise. This is an emerging association carried by Michael Robinson, Dan Moor and Aaron Cort, repeated cyber investments and current operator cases. The territory remains contested.

The first association is supported by behaviour. The second has durable external transmission. The third is gaining evidence and has yet to become a clearly owned market position.

1. Craft's operator claim has costly proof that its front door understates

Dan Moor's cyber GTM guide is the clearest example. It converts roughly 16 months embedded at Horizon3.ai into cohort analysis, channel design, win-rate evidence and a reusable Difficulty Ratio. Aaron Cort's Replit and Supabase pieces describe interim operating work across ICP selection, lifecycle, onboarding, launches and community. Cassie Chao Leemans adds a distinct talent system with hiring thresholds, assignments and structured judgement. Six polished founder videos use intimate interviews, office and product footage, and co-branded end cards to make that support legible through founder testimony.

This matters because many firms call themselves operators. Craft can point to named people, duration, interventions and operating artefacts. The limitation is architectural. Sixteen of the 24 accessible in-window articles are investment announcements and only six are Craft Insights. The homepage language could belong to many venture firms. A prospective founder has to discover the stronger proof article by article.

The apparent business loop is credible: embedded support produces methods; methods become articles, founder films and partner posts; LinkedIn and private networks create encounters; councils, event applications and operator relationships create introductions; portfolio outcomes return as proof. The final connections remain unverified. Public evidence does not show how many qualified founder relationships, customer introductions or investments the content created.

2. The most owned historical idea is no longer the visible organising system

Burn Multiple and SaaSGrid show what Craft can own when a useful operator idea becomes infrastructure. Burn Multiple still travels through independent SaaS guides and a 2026 adaptation for direct-to-consumer businesses. SaaSGrid joined public education, a diagnostic tool and a fast investment invitation, then became a standalone company. That is a stronger observable content-to-business loop than the current feed.

The present system still contains the same analytical instinct, especially in the Replit, Supabase and Horizon3.ai work. It lacks an equivalent umbrella, cadence or destination. Bessemer's recurring cloud and AI franchise and First Round Review are easier to recognise and revisit. Craft appears to be migrating from SaaS-efficiency authority towards AI-era growth and security. That may be strategically sound. The current public architecture has not yet made the migration memorable.

3. The institutional voice is becoming less dependent on David Sacks

Laurie's 2025 hypothesis that prominent partners can outgrow their funds is directionally supported in Sacks's case. All-In and his public-policy role give him far more personal salience than Craft's owned channels, often with a destination and subject matter separate from the firm. Independent coverage also brings reputational scrutiny that Craft cannot control.

The current byline pattern offers a meaningful counterweight. Michael Robinson appears on 13 of 24 accessible in-window articles and now leads the investment team. Moor carries cyber GTM, Cort carries growth, and Leemans carries talent. Their arguments are closer to Craft's service proposition than Sacks's broad media presence. Craft is developing an institutional federation of specialist voices. Recognition still clusters around individuals, and the system needs stronger connective tissue to transfer that recognition to the house.

4. Unprompted is the clearest current attempt to turn expertise into a relationship market

Unprompted assembles Sacks, Robinson and Cort with Replit and Supabase leaders for an invite-only marketing and growth summit. It links current proof, a referral-only network of more than 200 senior operators, partner distribution and an application route. This is coherent with Craft's actual strength: selective practitioner relationships built around work already done.

As of 23 September, the 1 October 2026 event had not happened. Delivery, reuse, attendee quality and commercial consequence are unknown. It should be judged as programme architecture rather than demonstrated effectiveness.

Overall judgement

Craft's system is partly effective for operator credibility, founder sourcing and portfolio support. Proof is strong in selected cases; historical idea transmission is unusually strong; current firm recognition and association are weaker; relationship creation is plausible; content-attributed action and commercial outcomes are unproven. The strategic opportunity is to organise the existing evidence into a named, recurring system that makes Craft's embedded work recognisable before a founder already knows the people.

This reading should change if target-audience research shows a strong spontaneous firm-level association with operator deployment; CRM evidence attributes material founder, customer or hiring relationships to the content system; Unprompted becomes a recurring programme with reused outputs and verified outcomes; private network evidence shows the public layer is intentionally secondary; or complete platform analytics reveal coordinated distribution and stronger institutional reach than the accessible sample.