Jack Chan

The best way to win the debate over Chinese AI models is to change the decision. Ask the ban side: Can the U.S. really stop open model files from being copied? Probably not. Ask the no-ban side: Would you let an untested model run inside the Pentagon or the power grid? Probably not. Now both sides have given up their extreme position. The real decision is no longer “ban or don’t ban.” It becomes: Where should these models be allowed, and what should they be allowed to access? Keep normal use open. Put strict checks around government, defense, and critical infrastructure. That’s the persuasion lesson: When a debate gets stuck between two bad choices, don’t argue harder. Change the question to one people can actually decide.

In an new report expected to be released later today the Trump administration will argue that directing federal funding towards individual researchers, and research that uses AI, will accelerate scientific progress faster than funding collages and universites.

The best way to win the debate over Chinese AI models is to change the decision. Ask the ban side: Can the U.S. really stop open model files from being copied? Probably not. Ask the no-ban side: Would you let an untested model run inside the Pentagon or the power grid? Probably not. Now both sides have given up their extreme position. The real decision is no longer “ban or don’t ban.” It becomes: Where should these models be allowed, and what should they be allowed to access? Keep normal use open. Put strict checks around government, defense, and critical infrastructure. That’s the persuasion lesson: When a debate gets stuck between two bad choices, don’t argue harder. Change the question to one people can actually decide.

We must NOT let our companies use these Chinese models to save a few bucks. OpenAI and Anthropic are correct. This is vital national security. Please read Bing West's just released Cat 5. I respect the Chinese people greatly but these companies are run by the PLA for heaven's sakes.

SITUATION BREWING: The Trump administration is considering restricting cutting-edge Chinese AI models, with momentum reviving after the launch of Kimi K3, per Axios.

When Replit Agent misunderstood a task and deleted SaaStr founder Jason Lemkin’s production database, Replit had two frames to choose from: “We have backups. The user can recover.” Or: “We gave the AI too much access.” The first frame defended the product. The second protected trust. Screenshots were already spreading, and Lemkin didn’t know how to restore the backup. So Amjad Masad opened with: “We messed up.” He admitted Replit lacked dev/prod separation, then shipped default isolation, production read-only access, and better recovery controls within two days. The response spread positively—and Lemkin kept using Replit Agent. Decision lesson: frame the problem around what you control, own the trust gap, and make the fix visible.

Everybody knows the first Andy Grove story. In 1985, Intel’s memory business was being crushed by Japanese manufacturers. Management kept debating how to save it: build a larger factory, develop better technology, find a specialized niche. Every proposal carried the same assumption: Intel had to remain a memory company. Grove asked Gordon Moore: “If we got kicked out and the board brought in a new CEO, what would he do?” Moore answered immediately: “He would get us out of memories.” Grove replied: “Why shouldn’t you and I walk out the door, come back, and do it ourselves?” Intel exited memory and made microprocessors its core business. That story is famous. Almost nobody knows the second Grove story. Twelve years later, Intel was already the dominant microprocessor company. The threat was now coming from below. AMD and Cyrix were selling cheaper, less powerful chips for low-cost computers. Grove invited Clayton Christensen to Intel to explain what disruptive innovation meant for the company. When Christensen arrived, Grove told him that his schedule had changed. He had ten minutes. “Tell us what your model means for Intel.” Christensen refused. He knew the theory of disruption. He did not know Intel well enough to prescribe its processor strategy. Instead, he told Grove about steel. Steel minimills had entered at the bottom of the market, producing rebar: cheap, undifferentiated and low-margin. The integrated steel companies were happy to leave. Each time they abandoned a low-margin segment, their profitability improved. Then the minimills improved. They moved from rebar into bars and rods, then structural steel, then sheet steel. The incumbents kept retreating toward more attractive customers until there was nowhere left to retreat. Christensen stopped there. Grove completed the analogy himself. Low-cost processors were Intel’s rebar. AMD and Cyrix were the minimills. If Intel abandoned the low end because the margins were unattractive, its competitors would gain volume, revenue and technical experience there. Eventually, they could move upmarket and attack Intel’s core. Intel went on to make a serious push into low-cost PCs with Celeron. “Rebar” became internal shorthand for the threat from below. The interesting part is not that Christensen was a good storyteller. It is that he refused to play the role Grove initially gave him: the outside expert who tells the CEO what to do. A direct recommendation—“Intel should launch a cheaper processor”—would have been easy to reject. Do you understand our costs? Our architecture? Our channels? Our roadmap? The conversation would have become a test of Christensen’s semiconductor expertise. He would have lost, correctly. So he moved the problem into an industry where Grove had nothing to defend. There was no Intel strategy in the steel story. No existing decision to justify. No executive identity at stake. Grove could examine the mechanism before confronting its implications. Christensen also chose a story with causal structure, not a decorative analogy: Enter through a low-margin segment. Let the incumbent retreat. Use the foothold to improve. Move upmarket. Repeat. Then he withheld the final mapping. That matters. Grove did not receive a recommendation from an academic. He reached a conclusion about Intel using a model he now understood. The judgment remained his. I’d call this move The Unfinished Analogy: Explain a mechanism in a distant domain. Make the causal sequence visible. Stop before applying it to the listener’s situation. Let them finish. One caveat: Christensen did not invent Celeron in a ten-minute conversation. Intel was already paying attention to low-cost PCs. What he gave Grove was a sharper model of the threat—and language that could travel through the organization. The mark of a useful strategic conversation is not that the adviser delivers a brilliant answer. It is that the decision-maker interrupts and starts explaining the answer back. From the book What is Your Problem by Thomas Wedell-Wedellsborg. Filed under Unfinished Analogy.