The original value loop
Dev Mode’s handoff model centres on a human developer inspecting a design, extracting implementation context and translating it into code.
The value is a more legible path from design intent to implementation. The developer receives that value at the point where design context becomes useful to a build. Repeated collaboration gives the tool a continuing role in the team’s workflow.
This teardown focuses on that product loop within Figma, rather than claiming to explain Figma’s entire company growth system.
What changes when agents build
The starting question was personal: I can go from an idea to Claude or Codex to working HTML without opening Dev Mode. What happens to Dev Mode’s job when that becomes part of the workflow?
In agentic development, the developer increasingly supervises output while an agent consumes implementation context. Some teams also begin in the codebase and may not maintain a parallel design file.
Where the loop weakens
Three tensions follow. Human inspection is no longer the only route to implementation. A separate handoff can become additional work for code-first teams. And faster generation can compound inconsistent components, states and interactions.
These are strategic hypotheses to investigate. They do not establish that every developer or team has changed behavior in the same way.
Strategic alternatives
A Figma-first approach can improve the context agents receive: semantic meaning, state logic and more efficient retrieval. A code-first approach can derive product structure from the repository and identify drift without requiring the team to recreate everything in Figma.
The strategic choice is where Dev Mode can retain value as the starting point of the build becomes less predictable.
The proposed bet
Shift Dev Mode toward Agent Experience (AX) while keeping humans able to inspect and supervise the resulting work.
For Figma-first teams, the proposal includes semantic context, state logic, token-efficient retrieval and AX-Diff. For code-first teams, it includes repository-derived product structure and detection of AI-generated product drift.
These capabilities are the proposed strategy in this case, not claims about shipped Figma features.
Expected business effect
The hypothesis is that maintaining alignment between intent and implementation can create recurring value beyond a single handoff. That could strengthen Dev Mode’s place in an ongoing build-and-review loop.
Adoption, retention, monetisation and expansion effects remain hypotheses. No numerical uplift is claimed.
Experiment
A proposed experiment would compare a normal agent-assisted build with a build using the richer context and review loop. Evaluate whether the proposed approach reduces missing states, inconsistent components and manual clarification.
Record review effort alongside output quality. More context is only useful if it helps the team reach a better result at an acceptable cost.
Countercase
Teams may already solve these problems through repository conventions, tests and their existing development tools. Adding a separate system could increase process without producing enough value.
The proposal earns its place only if the experiment shows a useful advantage over that baseline.
A new loop
The proposed loop is: product intent → useful agent context → generated implementation → human review → drift detection → corrected context.
The next Dev Mode could be the system that keeps UX and AX aligned from intent to shipped product. Figma does not need to control where building starts to test whether that is a valuable job to own.