Taste is not innate talent, it's accumulated care and intentionality, and AI speeding up reps doesn't shortcut the time investment required to build it.
Loredana draws on 20 years of classical piano training to argue that expertise requires embodied, sensory practice that can't be compressed. She reframes taste as 'care from that perspective': tasteful experiences are ones where the creator anticipated everything, the way a great host thinks through every detail of a party. Faster AI-assisted reps don't replace the time needed to develop that judgment.
taste-and-craft
When AI makes production trivially easy, authenticity becomes the scarce differentiator, and audiences actively disengage when they sense a work is purely AI-generated without human intent behind it.
She argues that as more software and content gets produced at high volume, people already consuming existing work will seek out even more authenticity. She cites research that when audiences realize something was created by AI alone (versus AI as a tool in service of a human's vision), they disengage, which is why she insists 'AI is the output' is the wrong mental model versus 'AI is a tool.'
taste-and-craft
Figma's Config 2026 releases are built around a single question: how will AI change what people are able to imagine, not just what they can produce faster.
She frames technology's historical role (writing invented for accounting evolving into Shakespeare, photography invented to capture reality evolving into imaginative image generation) as expanding human imagination rather than just efficiency. This framing drove Figma to ship both AI-forward features (a design agent that hands off control) and expressive ones (code as a creative medium, motion), spanning what looks like a wide gamut.
ai-design-tools
The right pattern for AI creative tools is 'AI gets you to 70%, then you make it yours' - AI should set up parameters and a starting point, not attempt to fully realize a vision from a prompt.
Loredana says describing a fully-formed design in words is nearly impossible ('dictating a painting over the phone'), so Figma's shader-effects and motion tools let AI generate an initial pass with tunable parameters (speed, responsiveness, gradient values) that the designer then directly manipulates. This preserves direct manipulation as the finishing step rather than iterative prompt-and-critique.
ai-design-tools
Designers are increasingly building reusable systems and workflows rather than one-off screens, and Weave's node-based 'mini apps' are Figma's bet on this shift.
Weave, a company Figma acquired, lets users chain prompts, model inputs, and variables on an infinite canvas to produce a 'mini app' whose output always matches the brand and intent baked into the underlying workflow, regardless of what input is fed in. Figma is bringing this into the core product at Config, treating the workflow itself as the design artifact rather than any single output.
ai-design-tools
Review, not creation, is now the bottleneck across design and engineering, because AI has made producing work fast but reviewing it still requires human judgment humans don't naturally enjoy applying.
She notes people love making but don't love reviewing, and organizations are now flooded with AI-accelerated output that needs to be vetted. Figma's answer is code layers on canvas: instead of routing feedback through recorded Loom videos, teams can visualize a prototype's states and flows directly on the shared canvas, comment on specific screens, and clear the backlog collaboratively.
ai-design-tools
Evals are becoming core design work because design decisions increasingly run through non-deterministic, LLM-driven systems that need an explicit definition of 'good.'
Comparing this to her Meta experience where feed-algorithm decisions mattered more than button design, she argues that when the experience is produced by an LLM rather than a fixed algorithm, designers must specify quality criteria (evals) that a system or human reviewer can check against, even though design has no single correct answer, only many incorrect ones.
ai-design-tools
Roles between designers and engineers are blurring in both directions: designers ship PRs, and developers increasingly report doing more design work and rating design as more important.
Loredana says she personally ships PRs (mostly eval-tooling and small UI bug fixes) and had to figure out who should approve them, landing on asking the relevant VP of engineering. She cites an internal Figma study where developers, not designers, were the group most likely to say design is more important and that they are doing more of it themselves, suggesting designers may be underestimating how much design work now happens outside their own discipline.
human-ai-collaboration
Data and workflow synthesis across tools (Slack, task trackers, shipped product) is becoming a priority so designers have enough context to act with precision, not just speed.
She frames three layered needs revealed by internal research: better synthesis and context flowing into the system, the power to mold outputs with both speed and precision, and a way to combine everyone's individually AI-accelerated work into something coherent rather than just producing more disconnected output.
design-org-strategy
Figma's product strategy at Config was driven by collapsing the sequential, lossy handoffs between separate tools (ideation, design, code) into one interoperable canvas.
She describes the old flow as a chain of translations, one tool producing an artifact that becomes the brief for the next tool, with translation cost and no quick loop back to test a new idea at each step. Decisions like putting code layers and animation directly on the canvas are meant to remove the distance between these previously sequenced actions.
design-org-strategy
When hiring designers now, she prioritizes range, curiosity about new tools, and systems-level thinking over prior experience in a specific tool or domain.
She explicitly pulls back from filtering for candidates who worked specifically on design tools, instead looking for people who can find intersections between previously disconnected domains (design and code, sound and visual, PM and design) and who show passion, because passion drives the persistence and taste needed to keep learning a fast-changing toolset.
hiring-and-talent
A strong signal in hiring is whether a candidate is comfortable exploring the current state of AI tools without becoming purely 'AI-pilled,' since real value comes from finding the intersection between tool capability and human judgment.
She distinguishes between designers who explore what AI can do for them and designers who look for nothing but AI to do the work; she wants the former, people willing to develop range across a widening toolset the way a professional painter (versus a child with three brushes) needs many distinct tools, each suited to a different job.
hiring-and-talent
Companies
Figma - Loredana is Chief Design Officer; the episode covers everything the company shipped at Config 2026.
Meta - Loredana's previous employer for close to a decade before joining Figma about 10 months prior to this episode; contrasted for its focus on global-scale content versus Figma's maker culture.
Weave - A design product Figma acquired the prior year that uses a node-based interface on an infinite canvas to string together prompts and model inputs into reusable 'mini apps'; being brought into Figma at Config.
Paper - Mid-episode sponsor read; a tool that now lets users copy and paste layers, properties, images, and SVGs directly from Figma into its canvas.
Framer - Mid-episode sponsor read for Framer 3.0, pitched for letting the host run AI agents directly in a canvas to manage CMS and ship production sites.
Hey Marvin - Cited as an example tool for synthesizing user and product data (Slack, Asana tasks, shipped product) to help designers understand how users actually use a system.
Techniques and frameworks
Taste equals care - Loredana's reframing of 'taste' as intentional, anticipatory effort (her party-throwing metaphor) rather than an innate gift, and argues it's built through sustained, loved practice, not shortcutted by AI reps.
AI gets you to 70%, you make it yours - Her framework for how Figma's AI features (shader effects, motion) should work: AI sets up the workspace and parameters, then the designer takes direct manipulation control to finish it, because describing a fully-formed vision in words is far slower than adjusting it by hand.
Systems, not screens - Her framing that designers are increasingly building reusable systems/workflows (as in Weave's node-based 'mini apps') rather than one-off screens, where the underlying workflow bakes in the intent so any input still produces on-brand output.
Evals as the designer's job - Her argument that in non-deterministic, LLM-driven products, specifying and grading 'what good looks like' (evals) is now core design work, since design has no single correct answer but many incorrect ones.
Code layers on canvas - Figma's Config 2026 approach to collapsing the review bottleneck: bringing prototypes and their different states/flows onto the shared canvas instead of routing feedback through recorded Loom videos.
Summary
Loredana Crisan, Chief Design Officer at Figma, joins host Vid roughly ten months into her tenure to unpack everything Figma shipped at Config 2026 and the philosophy behind it. She opens with her background as a classically trained pianist turned sound engineer turned accidental startup designer, using two decades of piano practice to argue that taste is not an innate gift but accumulated care: the same intentionality a great host brings to a party, considering every detail from where guests hang their coats to the food. Pushed by the host's "design Twitter skeptic" framing that AI-accelerated reps might make deep experience less defensible, she holds firm that expertise still requires time and love of the craft, and that AI reps don't substitute for that investment.
The conversation moves to how Figma is operationalizing this philosophy in product. Loredana frames the entire Config 2026 release around one question: how will AI change what people are able to imagine, not just what they can produce. Features span from AI-forward (a design agent, shader effects and motion generation) to expressive (code as a creative medium), unified by a pattern she calls "AI gets you to 70%, you make it yours": AI sets up parameters and a starting point, but the designer takes direct manipulation control to finish it, because verbally describing a fully-realized design is nearly impossible. She connects this to Weave, an acquired product now integrated into Figma that lets designers chain prompts and model inputs on an infinite canvas into reusable "mini apps," reflecting her broader claim that designers increasingly build systems, not screens.
A significant stretch covers organizational pain points Figma is hearing from customers and addressing internally: review has become the bottleneck, not creation, because AI makes output easy to generate but hard to vet, and humans don't naturally enjoy reviewing. Figma's answer is bringing code prototypes onto the shared canvas so teams can visualize states and flows together instead of routing feedback through Loom videos. She also introduces "evals" as newly essential design work: when experiences are produced by non-deterministic LLMs rather than fixed algorithms, designers must explicitly specify what "good" looks like, even in a domain with no single correct answer.
Loredana describes blurring boundaries between designers and engineers, noting she personally ships small PRs (mostly eval tooling and UI bug fixes) and had to negotiate who reviews them. She cites an internal Figma study finding developers, not designers, are most likely to say design is more important and that they're doing more of it, a sign designers may underestimate how much design work is migrating elsewhere. Comparing Figma's maker-obsessed culture to her nearly decade-long run at Meta, she attributes Figma's product strategy debates (should code live on canvas, how should AI agents be represented, how do product surfaces stay interoperable) to a deliberate effort to collapse the lossy, sequential handoffs between separate tools into one connected canvas.
Closing on hiring, she says she looks past whether a candidate has worked specifically in design tools and instead prioritizes range, curiosity about the expanding toolset, and passion, since passion sustains the persistence and taste development that ultimately produces good work. Asked what she'd put on a billboard for the entire design community, she returns to her opening metaphor from a Bjork quote about music having no soul unless someone puts it there: her message is to put yourself, your soul, into what you make, regardless of the tool.
Notable Quotes
"You can't blame the computer if the music has no soul. If the music has no soul, it's because nobody put it there. That is the case with AI as well. AI is a tool, but your responsibility is to bring the heart to it." - Loredana Crisan
"Taste is a shorthand for like this person has it and they're miraculously somehow were born with it. It's very very much not true... for me taste equals care from that perspective." - Loredana Crisan
"When people realize that what's being put in front of them is inauthentic and it's just created by AI, they disengage. Research says this very very clearly." - Loredana Crisan
"Review is a bottleneck across design, across development right now... we actually love making and we don't love reviewing quite as much. But we're now in a place where we have to become expert reviewers." - Loredana Crisan
"Design is having an identity crisis and a renaissance at the same time." - Loredana Crisan