Figma CEO Dylan Field: AI Makes Great Thinkers More Valuable, Not Cheaper
Figma CEO Dylan Field stated that artificial intelligence makes great thinkers more valuable, not cheaper. He believes the real advantage lies in second and third-order thinking, not in the initial answers provided by models.
He described the corporate response as a loop: first being surprised by what models can do, then emerging from the other side. Some companies completed this loop by the end of last year, while others are just starting now. He did not name which companies have completed it or provide a standard for completion.
After exiting the loop, his judgment is that companies still have employees, and people still have jobs. Figma aims to retain great thinkers because he wants to hear their second and third-order judgments, rather than taking the first generated results as decisions.
He pointed out the risk of letting AI think for people and then throwing the results over the wall. He believes a better use is to learn faster with it, try things that were previously impossible, and thus be bolder. Thinking remains the responsibility of humans, while models are responsible for expanding the range of experimentation.
This is not a guide for layoffs or hiring. The buyers are companies still going through the loop, while the sellers are models, tools, and those who can still make second-order judgments. The event is driven by the sudden visibility of model capabilities: first lowering the price of initial outputs, then marking up those who can review, compare, and redirect. Beneficiaries are teams treating AI as a testing ground, while those under pressure are roles and processes that only forward the first output.
Public statements did not provide Figma's hiring numbers, salary changes, or model procurement amounts. Three points can be verified: the loop has a sequence, employees have not been declared to disappear, and the danger is defined as outsourcing thinking and throwing it over the wall.
Source: Public Information
ABAB AI Insight
Dylan Field had already stated internally in October 2026 the same rule: Figma hires people because they can think, and not to outsource thinking to AI. Externally, he also requires that if AI is used in the writing process, it should be noted at the beginning. This aligns with his path in 2012 when he received a $100,000 Thiel Fellowship and left Brown University with Evan Wallace to create Figma: tools first lower production costs, then differentiation comes from judgment. In 2022, Adobe's acquisition of Figma for about $20 billion was blocked by regulators, and in July 2025, Figma went public at an issue price of $33, closing its first day at $115.50, indicating that the market values collaborative judgment rather than single-instance outputs.
Capital has not been directed towards building foundational models. Index Ventures, Greylock, Kleiner Perkins, and Sequoia hold stakes in the canvas and collaboration layers; Field's personal angel investments are more in tools like OpenSea and Loom. In 2026, he sent products from Claude Code back to editable layers via Figma MCP, then pulled canvas modifications back to the codebase. The motivation is that generation has become cheap, while what is scarce is juxtaposing solutions, cutting off bad directions, and deciding which version enters the product. Second-order thinking is placed on a paid interface, rather than on model calls.
This is similar to how Bloomberg terminals did not disappear due to data digitization, and it is close to Adobe's position after desktop licensing was disrupted by browser collaboration. Canva captures the template scale, foundational model companies capture the first generation, and Figma positions itself in the comparison and rewriting after generation. The industry phase is about control, not building models from scratch: those still surprised by what models can do are still in the first half of the loop.
The essence is the transfer of pricing power. After models push the marginal cost of first-order outputs towards zero, the payment point shifts from "who can produce it" to "who can judge whether it should be done and the consequences of the next layer." Throwing results over the wall means giving pricing power to model suppliers and the process itself; using models to expand experimentation keeps pricing power with those who can make second and third-order choices. Employees have not disappeared; what has disappeared is the pricing for those who only produce first-order outputs.
ABAB News · Cognitive Laws
- The cheaper the first-order output, the more expensive the second-order judgment.
- Tools are responsible for boldness, while humans are responsible for the consequences.
- Only after completing the loop do we find that positions still exist, but pricing has changed.