ExpertsFMThe New Full-Stack Designer
Y Combinator Startup Podcast · October 28, 2026
5 Selected Lessons:
framework · 2 reasoning_pattern · 1 heuristic · 1 priority · 1
Reduction/curation is the hard, valuable part of design work
Source
Y Combinator Startup Podcast — The New Full-Stack Designer (October 28, 2026) · at 48:47 ▸
When this applies
when a team must decide what to cut from a product, page, or pitch that has many stakeholders wanting their piece included
The takeaway
Rank curation and editing above addition as the core, harder skill of good design.
Why it works
Adding everything is easy because it avoids conflict; removing requires difficult internal negotiation about what matters most, which is why edited work is rarer and signals more effort/care than packed work.
One level up
The visible simplicity of a good result often hides a harder internal process of saying no.
Where it stops holding
This stops working when stakeholders genuinely need comprehensive access to all information (e.g., technical documentation reference pages) rather than a persuasive narrative.
Detail
hierarchy:
- Decide what to cut (hardest, most valuable)
- Space and sequence what remains
- Add polish/visual craft last
rationale: Everyone's competing, everyone wants their functionalities or features on the page — reduction requires harder internal conversations but pays off.
In their words
It's easy to throw everything. It's much harder to figure out what needs to be edited out. Because that conversation is just much harder to have with the team. Everyone's competing, everyone wants their functionalities or their features to be on the page.
Progressive disclosure / 'Primal' content system
Source
Y Combinator Startup Podcast — The New Full-Stack Designer (October 28, 2026) · at 5:54 ▸ · at 1:02:01 ▸
When this applies
when presenting a lot of information to a user who could easily be overwhelmed
The takeaway
Reveal information in layers tied to user intent, never all at once.
Why it works
Chunking content into discrete, one-layer-away chapters lets users explore voluntarily instead of being bombarded, which keeps engagement comfortable and self-directed.
One level up
Respect the user's attention as a scarce resource you must earn access to, not demand.
Where it stops holding
This stops working when users need all information immediately to complete a task (e.g., safety-critical or transactional flows) where hiding content costs more than it saves.
Detail
procedure:
- Identify core topics (e.g., technology, news, careers)
- Treat each as a homepage 'chapter' one layer deep
- Let users explore via progressive disclosure instead of navigation menus
- Use sheets/transitions to go deeper without losing homepage context
rationale: Users feel they are exploring the site rather than the site demanding their attention.
In their words
Everything is just one layer away and everything is progressively disclosed. So you never overwhelm the user with too much content in general.
Decompose complexity into sub-tasks for AI agents
Source
Y Combinator Startup Podcast — The New Full-Stack Designer (October 28, 2026) · at 19:50 ▸ · at 23:06 ▸
When this applies
when directing a capable but non-expert AI tool on a complex multi-part creative/technical task
The takeaway
Decompose a complex goal into explicit, checkable components before handing it to an AI tool, and reassemble iteratively.
Why it works
AI executes well on well-specified sub-problems but struggles with an underspecified whole; breaking a complex target into explicit components (lighting, shadow, structure) gives the model the missing context it needs.
One level up
The quality of AI output is bounded by the clarity of the human's own understanding of the problem's structure.
Where it stops holding
This stops working when the operator themselves doesn't understand the underlying components well enough to decompose the task meaningfully.
Detail
procedure:
- Identify the full concept (e.g., realistic grass shader)
- Break it into components (lighting source, shadows, image parts, underlying terrain structure)
- Feed components individually to the agent
- Reassemble into the whole
- Use a second AI agent to interrogate/clarify process gaps
rationale: AI is extremely capable of execution but needs the complexity pre-decomposed by someone who understands it.
In their words
I think, number one, Codex and all the current AI software and the models are extremely capable at creating anything. You can use these tools to make anything that you want, but you have to know what you're doing. And you have to be able to take a concept or something more complicated
Designer-as-editor in an AI workflow
Source
Y Combinator Startup Podcast — The New Full-Stack Designer (October 28, 2026) · at 0:00 ▸
When this applies
when a capable generative tool can produce raw output faster than a human, but has no taste
The takeaway
When a tool removes the bottleneck of production, the bottleneck shifts entirely to judgment and curation.
Why it works
AI can execute complex technical output but has no sense of what looks or feels good, so a human must continuously select, reject, and redirect it.
One level up
As execution becomes cheap, taste becomes the scarce and valuable skill.
Where it stops holding
This stops working when the human curator lacks a clear enough bar or vocabulary to evaluate outputs, in which case AI just produces noise faster.
Detail
approach: Break a complex design goal into sub-tasks, feed them to AI, review output, and iterate with explicit quality judgments rather than accepting first output.
signals:
- output volume is no longer the constraint
- role shifts from making to selecting
- continuous back-and-forth with the agent
rationale: AI raised the ceiling of what's executable but can't judge aesthetic quality itself.
In their words
100%. I mean, AI has definitely raised the ceiling for all of us, but it's not a crafter. It can execute complexity and technology that I couldn't handwrite code. It can write code, but it doesn't know what looks good and what looks bad.
Micro-interaction craft as a trust signal
Source
Y Combinator Startup Podcast — The New Full-Stack Designer (October 28, 2026) · at 8:56 ▸
When this applies
when a product's audience is itself expert in the thing being sold
The takeaway
Invest disproportionate care in details your most discerning customers will specifically test for.
Why it works
When your buyers are themselves craftspeople, small details signal the level of care you'll bring to their paid work, so detail becomes a proof point rather than decoration.
One level up
Credibility with experts is earned through demonstrated craft in exactly the dimension they judge you on.
Where it stops holding
This stops working when the audience is not expert enough to notice or value the detail, making the investment unrecognized and wasteful.
Detail
situation: Collins, a branding agency, needed a website that demonstrated design craft to potential clients who are themselves designers.
decision: Dedicate an engineer's focused time to hand-craft one small bounce interaction rather than using an off-the-shelf tab/animation.
outcome: The detail became an 'Easter egg' that signaled the level of craft clients could expect if they hired Collins.
lesson: For audiences who judge craft, invest where they'll specifically look for it.
In their words
I think the clients coming to the website are looking for those kind of Easter eggs. They want to know that if I hire this agency or if this agency takes this much detail to produce a website of that quality, what kind of Easter eggs are they going to put into our projects?