David Friedberg: After Building a CRM Over the Weekend, Still Opted for Salesforce
Ohalo Genetics CEO David Friedberg stated during the All-In podcast and Salesforce earnings call that the company built an internal customer relationship management (CRM) system over a weekend using Claude Code and Cursor, but later decided to purchase Salesforce. He mentioned that maintaining reliability, accounts, security, permissions, and data protection becomes significantly more complex than a weekend prototype when scaling up.
He prioritized time and funds for proprietary workflows such as plant breeding software and farm customer development, believing that these are the differentiators. There is no need to reinvent cross-industry infrastructure: there is no need to create another Slack, Gmail, or Salesforce, and no one wants to reinvent Excel. Employee Ben integrated Salesforce with Claude and Cursor, completing customizations that had been stalled on another CRM for months in less than a month, leading him to conclude that this was the better approach.
Friedberg revised his earlier statement that "SaaS is dead": system-recording platforms will not be replaced by weekend-generated programs all at once; the real pressure is on thin-functionality vertical SaaS that serves only a single industry, as companies better understand their own processes and can more easily redo them with generative tools. Salesforce's stock surged after the earnings report, with CEO Marc Benioff incorporating this customer story into the narrative of "no SaaS apocalypse" and launching a pilot plugin to embed CRM into Claude with Anthropic.
Ohalo also runs communications on Slack. The self-built prototype addressed fields and pages, while purchasing an existing platform brought in auditing, identity, and ongoing patches. An agent scanned county-level public databases to find farmers' planting and contact information, which was placed in a custom interface on Salesforce rather than in a weekend warehouse.
Generative coding transformed the question of "can we build a CRM" from a quarterly project into a weekend demo, while leaving the question of "can it serve as a production system" tied to compliance and operational costs. Horizontal platforms charge rent for infrastructure shared across multiple industries, while vertical tools charge for industry knowledge rent, which is easier to rewrite internally.
In market mechanisms, buyers are agricultural tech companies that want to invest R&D in breeding rather than forms, while sellers create annual fee platforms for identity, permissions, and record systems. The event was triggered by cross-references from the earnings report and podcast. Funding shifted from self-built engineer hours back to Salesforce subscriptions, and vertical SaaS budgets were marked as replaceable. Beneficiaries are the system-recording layers and giants that can be embedded into agents, while those with narrow-function subscriptions that can be clearly replicated in a weekend are under pressure.
Source: Public Information
ABAB AI Insight
Friedberg previously labeled vertical software as the worst asset class, arguing that engineers could use generative tools to redo high-priced industry suites. This time, he testified during Benioff's earnings call: a weekend CRM can be showcased but cannot be relied upon. The SaaS apocalypse narrative was downgraded to lowercase, targeting thin verticals while sparing cross-industry operating systems like Slack and Excel. After Climate Corporation was sold to Monsanto, he recognized that the moat for agricultural software lies in field data and breeding models, not in creating another lead dashboard.
The capital path is: first proving "can do" with Claude and Cursor, then using the same tools to establish Salesforce within a month, avoiding locking core engineers in a permissions matrix. Benioff needs a narrative that shifts from skepticism to payment, while Friedberg needs to align R&D focus on breeding. Removing the CRM screen and embedding skills into conversations signifies the platform's acknowledgment that interfaces can be replaced by agents, but the recording layer cannot. Stocks rose due to this alignment, as the purchase remains for system records.
Similar structures are seen in companies continuing to use Workday for payroll, NetSuite for accounting, and Excel for analysis: generative tools improve form speed but do not alter the legal implications of audit trails. Vertical SaaS is under pressure to be rewritten internally, while horizontal platforms are in an expansion phase treating agents as new clients. What is truly being replaced are products that lack multi-tenant security stacks beyond industry terminology.
The essence is that technological substitution delineates boundaries. The mechanism is: generative coding reduces the replication costs of interfaces and workflows, while increasing but not eliminating the fixed costs of identity, permissions, and data residency; companies outsource fixed costs to cross-industry platforms and retain variable industry knowledge for self-building. Those who only sell thin processes stand on the side that can be replaced by weekend projects.
ABAB News · Cognitive Laws
- Systems that can be completed over a weekend often still lack operational support and auditing.
- No one wants to reinvent Excel; what they want to recreate is the layer with only industry terminology.
- Core competitiveness lies not in generic forms, but in processes that others cannot replicate over a weekend.