How To Design In The Agent Era

Predictions and Expectations:

• Stephen Haney expects that models will eventually get better at tactical design decisions like letter spacing over time.
• Haney believes that while models will improve at tactics, they are "probably 10 years" away from being able to make high-order design decisions regarding value propositions or market needs.
• Haney expects that the design tooling will accelerate to keep the quality bar up at the same pace as engineering and product teams are shipping software.
• Haney expects that there will be more design need in the world because there are more companies and more software being created.
• Haney predicts that the role of design in organizations will not change significantly as a function of exploring problem space or dealing with stakeholders.
• Stephen Haney expects that agents are currently not ready for the precision required to build complex design tools at the level of Figma.
• Haney believes that competitors who have every single feature but lack attention to detail are at a disadvantage because people want Paper to be better.
• Haney expects that eventually, every company will go upmarket and stop being "underground" or cool when they start selling to enterprises.
• Stephen Haney predicts that models will eventually get good at "tactics" like not making everything bold.
• Haney expects that trends will eventually take over even before models become dominant, noting that everyone looks like Linear for a couple of years.
• Haney expects that humans will remain an essential piece of building stuff that stands out from the crowd.
• Stephen Haney expects that agents will get "smarter and smarter" over time but currently lack skills in organizational decision-making.
• Haney expects that agents are already doing more work than anticipated, noting it is a challenge for the company to scale servers to meet agent demand.
• Haney expects that the "cloud code explosion" of December 2025 will drive designers onto the desktop and into local repositories.
• Haney expects that the design tool will become part of the entire stack rather than existing in a silo, removing the "wall" between design and code.
• Stephen Haney expects that the future of design involves a new curation process where agents generate hundreds of variations overnight for humans to curate.
• Haney expects that "one person can do the work of an entire team" using tools like Paper.
• Haney expects that the "best product, most authentic product wins" for design software, even if the product is missing features early on.

Timelines and Milestones:

• Stephen Haney plans to launch the "Paper Mono" font in "a couple of weeks."
• Haney states the company is planning to work on organizing prototypes and feedback cycles over the "next six months or so."
• Haney references the "cloud code explosion" happening in "last december 2025" as a past milestone that shifted their strategy to the desktop.
• Haney mentions having "a couple of years" between selling his first company and starting Paper.
• Haney notes that trends like the "Linear look" tend to dominate for "a couple of years."

Technology and Product Direction:

• Stephen Haney plans to build out "table stakes" features including components and comments to address current bottlenecks.
• Haney plans to add a feature that allows users to iFrame prototypes onto the Paper canvas to organize them and leave comments.
• Haney expects the product to shift focus toward "agentic workflows" where tools speak well to agents.
• Haney plans to continue building the open-source "Paper Shaders" library as a form of marketing rather than product development.
• Haney expects that agents will eventually be able to "read the CSS out" and work with Paper files directly without translation.
• Haney plans to use MCP (Model Context Protocol) servers in the desktop app to allow agents to run tool calls like "write HTML."
• Haney expects the product to allow agents to "pull designs back into the code base" and match existing coding styles and conventions.
• Haney plans to eventually build "components" in Paper, though they are currently not available.
• Haney expects the product to handle "design to code" and "code to design" workflows seamlessly without format switching.
• Haney plans to integrate directly with tools like Cursor and Conductor to form a "new agent stack."

Market and Industry Outlook:

• Stephen Haney expects that the market for design tools will see Paper making a dent, noting it is the first company since Figma to do so.
• Haney expects that "vibe coded" apps with light/dark modes, purple gradients, and small headers are common but often look "sloppy" and "average."
• Haney expects that "AI tells" like overusing bold fonts, too many font sizes (5-8 instead of 3), and overusing cards are common mistakes in current AI-generated designs.
• Haney expects that great companies of the "last 10, 20 years" have all had exceptional design.
• Haney expects that mediocre presentation can negatively impact hiring and the talent a company can attract.
• Stephen Haney expects that agents are "insecure" and tend to fill space with unnecessary icons and numbers.
• Haney expects that the "average thing" is not exceptional and people are not excited to use it.
• Haney expects that the market will forgive a product missing features if the company is authentic and values-aligned.
• Stephen Haney expects that Figma has moved upmarket and is no longer "underground at the cool events."
• Haney expects that the "vibe coded" look is becoming a standard that designers need to actively avoid to stand out.
• Haney expects that the "four horsemen of the apocalypse" (bad design trends) include all caps with extra letter spacing and badges with extra little icons.

Company Plans:

• Stephen Haney plans to set up a "merch shop" for Paper merch, including shirts like the "Paper Monos" shirt.
• Haney plans to release a "mini site" for the Paper Mono font soon.
• Haney plans to maintain a team of 12 people (designers and engineers) to keep the talent bar high.
• Haney plans to continue talking to a designer "every single day" as a core company philosophy.
• Haney plans to have all employees talk to users in Discord or Slack to stay informed.
• Haney expects to continue using "Cloud Code" and "Cursor" extensively for coding agents internally.
• Haney plans to continue building the brand by releasing open-source tools like the shader library and fonts.
• Haney plans to focus on "bottlenecks" first, currently moving from early adopter acquisition to feature development (comments, components).
• Haney plans to build the "best design tool possible for this next era" out of love for design.
• Haney expects that the team will continue to "live our values" as a company of 12.
• Haney plans to not use agents in their internal Slack but to use them for coding and marketing videos.
• Haney plans to have engineers read "every single line" of code, sometimes multiple times, to ensure quality.

Financial Guidance:

• No financial guidance, revenue projections, or specific budget figures were mentioned regarding future performance.

Risks and Caveats:

• Stephen Haney notes that it is "very tempting" to use whatever cloud design AI spits out, but warns it "doesn't help you stand out from the crowd."
• Haney cautions that "models just aren't there yet" for the precision required to build a complex design tool.
• Haney warns that "bad dark mode" (using literal black or reversing colors) is a common "AI tell" that makes designs look sloppy.
• Haney expects that "vibe coded" designs often lack trust, which is critical for brands dealing with finances.
• Haney warns that using too many font sizes or bold styles makes designs look like "one of a million other projects."
• Haney cautions that overusing icons, pills, and numbers can make a design feel like "an insecure designer."
• Haney warns that the "average thing" is not something people are excited to use.
• Haney notes that models often "screw up contrast" and text readability.
• Haney warns that "widgets for widget's sake" that don't communicate value are a common AI mistake.
• Haney cautions that relying solely on AI-generated designs can make a company look "less intentional" and "less like you care."
• Haney warns that keeping two copies of a design system (design tool and code base) is a "tractable problem" that leads to accuracy issues.
• Haney notes that some AI-generated variations may "lose a little of the excitement" of the original design if not edited.
• Haney warns that "light mode and dark mode" are often unnecessary "MVVs" (minimum viable products) that models generate for free.

Confidence and Disagreement:

• Stephen Haney is "big believers" in the human element of design and how important it is compared to agents.
• Haney is "confident" that he will "never think [AI] is going to replace a human making these design decisions."
• Haney states he is "not sure" if models will reach human-level design decision-making in 10 years, calling himself "maybe too pessimistic."
• Haney is "excited" to see the Paper Mono font launch and the reaction from users.
• Haney expresses "fear" or concern that designers might use "vibe coded" designs without checking them, resulting in a lack of trust.
• Haney is "honored" that the Paper community cares so much about the product and wears the merch.
• Haney believes that "quality software takes time" and that moving slower can still result in a better product.
• Haney is "confident" that being values-aligned with the market will lead to forgiveness for missing product features.
• Haney is "not sure" if models will ever be able to fully verbalize the implicit thought process of a human designer.
• Haney is "convinced" that design is a "great differentiator" for great companies.
• Haney expects that the "next thing" that stands out will always require humans to "dry" (push) the limits.