OpenAI CEO Sam Altman: Large Companies Struggle to Maintain Cross-Department Agility
Someone from OpenAI's engineering side summarized the company's superpower as the ability to quickly assemble a group of authorized high-capacity individuals to tackle the most important and urgent tasks, with organizational boundaries being almost disregarded. This allows for flexibility, seizing opportunities, and recovering from mistakes. Sam Altman shared that startups are naturally good at this, while large companies find it difficult to maintain, which he believes is an underestimated area.
This statement frames the "cross-departmental task force" as a scarce capability after scaling, rather than just a cultural poster. Startups have fewer people and assume everyone is responsible for the same priority; as companies grow, planning cycles extend, safety and compliance meetings are held annually, and whether data can enter the system must go through committees, making cross-line mobilization an exception. Altman has repeatedly pointed out the inertia of large companies: technology changes every quarter, but organizations still make decisions annually.
OpenAI itself has become a large company, yet it still uses the phrase "no one cares about organizational lines" to explain how it can pivot between model and product windows. This statement serves as both an internal management declaration and an external observation of entrepreneurship: a large company that can turn task forces into repeatable actions will achieve the reaction speed typically associated with startups, without reverting to small teams. He noted that this area is still under-researched, indicating that the missing element is not slogans, but the institutional design of staffing, authorization, and prioritization.
He has long viewed speed as the primary advantage of startups over giants, stating that during technological upheavals like the internet, mobile, and artificial intelligence, the faster-moving side consumes the slower-planning side. During his time at Y Combinator, he tracked founder traits, associating the ability to quickly complete tasks with decision-making speed as nearly synonymous with success. This time, he elevates the same logic from individual founders to temporary internal task forces within companies.
In market mechanisms, this is the pricing of organizational capability, not product releases: the buyer is the project leader trying to seize talent during a window period, while the seller is the high-capacity individual being pulled from their original department. Funding manifests as computing power and salaries still tied to the original structure, but outputs are credited to the task force. The beneficiaries are platform-based AI companies that can break reporting lines; the pressured parties are traditional software giants locked into annual roadmaps. The event-driven nature is confirmed publicly by top management.
If large companies can make "assembling task forces" routine, the speed premium of startups will be partially reclaimed.
Source: Public Information
ABAB AI Insight
Altman shifts OpenAI's operational approach from model capabilities to personnel scheduling: superpower is not about having an extra department, but the ability to dismantle departmental walls at any given moment. This aligns with the founder advantages he observed at Y Combinator—small teams default to being responsible for the same urgent matters; after scaling, urgent matters must navigate budgets, safety, and reporting lines, causing windows to close. He acknowledges that he is managing a large company, yet frames "maintaining this startup essence" as a management category that has yet to be thoroughly researched.
The capital path is to use authorization instead of restructuring. Rather than opening new divisions, high-capacity individuals are temporarily assigned to the highest priorities, with computing power and salaries still linked to the original organization, while results are credited to the task force. The motivation is to iterate models on a weekly basis, avoiding annual planning. Resource mobilization relies on top management endorsement to reduce cross-line costs, rather than drawing a new organizational chart.
Analogies can be drawn to wartime project teams, Bezos's two-way door decision-making, and Google's past use of independent small teams to counter its own annual planning. The industry phase is that artificial intelligence has made "quarterly changes" the norm; if organizations still meet annually, creative destruction will hit from external startups to internal structures. The positional battle occurs within platforms: whoever can repeatedly assemble task forces will price like a startup during the window period.
Structural judgment belongs to the technical replacement at the organizational level: using a human network that can be restructured at any time to replace the monopoly of fixed departments over priorities. The mechanism is: when capability density is sufficient, the bottleneck becomes scheduling authority; if scheduling authority is still allocated along organizational lines, the speed advantage evaporates internally, leaving only external startups able to utilize it. The unexplored space is how to institutionalize this scheduling without relying on the founder's personal endorsement.
ABAB News · Cognitive Laws
The speed of startups is essentially the absence of organizational lines.
As companies grow, urgent matters must first go through committees.
Large companies that can repeatedly assemble task forces will reclaim startup premiums.