✨ Product & UX
Designing AI features users actually want. Latency, trust, streaming, citations, undo, the "AI moment" in a product.
The workflow
flowchart LR
A[User insight] --> B[Define AI moment<br/>where LLM helps]
B --> C[Design loop<br/>input → AI → review]
C --> D[Latency & UX<br/>streaming, skeletons]
D --> E[Trust signals<br/>citations, undo, edit]
E --> F[Measure adoption<br/>+ task success]
Trust > capability. Citations, undo, and visible thinking matter more than model size.
Key takeaways
Videos (55)
Beyond Components: Designing Generative UI for MCP Apps — Ruben Casas, Postman
Separate where an agent interface runs from how it is generated, and choose a UI approach that balances flexibility and predictability.
tldraw.computer - Steve Ruiz, tldraw
Spatial canvases like tldraw turn AI generation into composable graphs where drawing, annotation, and node-wiring become the programming interface.
Your Attention Is the Bottleneck, Not Your Agents — Zack Proser, WorkOS
Design agent workflows around sustainable human attention, with complete verification loops and filtered demands on your time.
Climbing the Ladder of Abstraction: Amelia Wattenberger
The future of AI interfaces isn't chatbots — it's structured UIs that automate small steps and let users zoom up and down a ladder of abstraction.
Design Patterns for AI Trust: Juries, Libraries, and Agent Tiers — Alex Bauer, Upside.tech
Build agent trust through clear intent, consistent business definitions, and shared guidance before expanding autonomous work.
Survive the AI Knife Fight: Building Products That Win — Brian Balfour, Reforge
Differentiation in AI products comes not from the model but from how you combine your proprietary data and functionality around commodity AI Lego blocks.
The End of the Static Screen: Architecting Intent-Driven UX — Gus Iwanaga, commercetools
Useful generative interfaces need design constraints and consistency in addition to correct tool orchestration.
Agentic Sites: Building Hyper Personalized Websites — Carlos Sanchez, Adobe
Real-time website personalization requires grounded content, controlled page components, and evaluation of both quality and latency.
Taste & Craft: A Conversation with Tuomas Artman, CTO Linear & Gergely Orosz, @pragmaticengineer
When AI makes shipping cheap, taste and disciplined product judgment — not speed — become the differentiating moat.
The End of Apps — Kitze, Sizzy.co
Personal agentic assistants built on Claude Code-style harnesses are replacing traditional productivity apps and SaaS.
Velocity Sickness: What Happens When Your Whole Team Gets 10x Faster — Matt Dailey, Ref.
Convert individual AI speed into team impact through coordinated planning, durable context, and clear ownership of important decisions.
Everyone Gets A Software Company — Benjamin Guo, Zo Computer
A persistent personal cloud with an agent can let nontechnical users build and operate an integrated software stack around their own needs.
ChatGPT is poorly designed. So I fixed it
You can patch ChatGPT's split-personality UX today with Realtime API + tool calls that route between voice, text and reasoning models on the fly.
Building the platform for agent coordination — Tom Moor, Linear
Linear's edge in the agent era is being the pragmatic, high-bar coordination layer where humans and AI agents share work, not a rushed copilot.
Why Your AI UX Is Broken (and It's Not the Model's Fault) — Mike Christensen, Ably
Treat agent sessions as durable shared resources to support reliable, interactive AI experiences.
Why your product needs an AI product manager, and why it should be you — James Lowe, i.AI
For AI products, resolve the AI capability uncertainty with evals and real-user tests before product building — then go wide on features and ruthlessly strip back.
The New Primitives: Building AI Native Software — Kwindla Kramer, Daily
AI-native software will require new interaction and computation abstractions beyond today’s agent loops.
Design like Karpathy is watching — Zeke Sikelianos, Replicate
Your docs' primary audience is now an LLM — ship llms.txt, curl snippets, and an MCP server, not pretty branded pages.
The Intelligent Interface: Sam Whitmore & Jason Yuan of New Computer
The next computer interface uses pose, audio, gesture, and tone as implicit signals so the device adapts to humans, not the reverse.
Imagination Engineering: "Live in the future and then build what's missing."
Capture and share your thinking so AI can help turn emerging ideas into concrete creative experiments.
From Arc to Dia: Lessons learned building AI Browsers – Samir Mody, The Browser Company of New York
Build prompt/eval tooling directly into your product so the whole team can iterate AI features with real user context — and use techniques like GEPA over RL.
The era of unbounded products: Designing for Multimodal IO: Ben Hylak
Win at AI product design by imposing app-specific structure — hierarchy, familiarity, surfacing — on top of the unbounded chat/multimodal canvas.
Everything is ugly, so go build something that isn't — Raiza Martin, Huxe (ex NotebookLM)
In the chaos of AI-augmented teams, individual taste and personal clarity become the durable engine for shipping non-generic, beautiful products.
The Death of Developer Advocates — Stephanie Jarmak, Sourcegraph
Agent advocacy turns tool-use traces and recommendation behavior into actionable product feedback.
Let's Talk About FOMAT: Fear of Missing Agent Time — Michael Richman, Cmd+Ctrl
Long-running coding agents need accessible notifications and session control wherever their users are.
Form factors for your new AI coworkers — Craig Wattrus, Flatfile
AI form factors range from invisible to conversational, and the designer's new job is character-coaching the model and crafting the right box around an inherently parallel collaborator.
How We Got LLMs to Recommend Our Open Source Library — Christopher Burns, Inth
Make accurate documentation easy for agents to discover both on the web and inside installed packages.
The UX of AI: Making AI-Powered Apps Your Users Don't Hate - Kathryn Grayson Nanz, Progress Software
Bridge users' existing mental models to AI capabilities through guidance and carefully adapted interaction patterns.
Feedback Loops are All You Need — Mehedi Hassan, Granola
Generic chat features fail in production — wrap your LLM in tight, user-driven feedback loops that constrain it to the workflows you ship.
Excalidraw: AI and Human Whiteboarding Partnership - Christopher Chedeau
Treat AI as a chance to rethink the product, not as a sprinkle on the old UI — and only ship AI features that match how users actually use your tool.
Good design hasn’t changed with AI — John Pham, SF Compute
Feature parity is no longer a differentiator in the AI era — speed, trust, accessibility, and delight are the durable design moats.
Second Order Effects of AI: Cheng Lou
Predict AI's second-order effects by asking who is doing the learning and where information bandwidth can be widened beyond current low-bandwidth interfaces.
The Missing Layer: Design Taste in AI Agents — Hassan El Mghari, Together AI
Give coding agents explicit design standards and strong references, then iterate deliberately on the interface.
Lessons from Studying Every Memory System — Shlok Khemani, Independent
Personalization needs both concise persistent context and retrieval, with mechanisms for correcting stale or mistaken memories.
Building AI Products That Actually Work — Ben Hylak (Raindrop), Sid Bendre (Oleve)
Build AI products by shipping, observing real production behavior and iterating — evals are necessary but they cannot tell you how good your product actually is.
The Prompt Is Still a Punch Card - Ted Johnson, JoinIn AI
Improving AI interaction requires changing the conversational protocol, not merely adding richer input channels.
500 people vibe-coded for 30 days. I was one of them. - Sanja Grbic, Automattic
AI adoption accelerates when organizations pair autonomy with practical training and colleagues who enable others.
Think You Can Build a Game with AI? Think Again! - Danielle An & David Hoe, Meta
AI makes game creation more accessible while increasing the importance of creative direction and playtesting.
Books reimagined: AI to create new experiences for things you know — Lukasz Gandecki, TheBrain.pro
Pair multiple AI primitives behind a polished UX and hide the AI itself—books with scene-matched music, avatars, and voice Q&A become a new medium.
On Curiosity — Sharif Shameem, Lexica
Building demos driven by curiosity is the primary way to discover what frontier models can actually do.
You Can't Prompt the Room: The Last Skill AI Won't Replace - Balázs Horváth, VisualLabs
The advantage in AI-assisted software development increasingly comes from understanding business needs and specifying valuable outcomes.
The Bitter Layout or: How I Learned to Love the Model Picker — Maximillian Piras, Yutori
Chat plus model picker is the AI-era equivalent of mode-laden UIs; design for the moving boundary between integrated and modular AI stacks rather than fixed model capabilities.
Chat and citations won't save your vertical AI - Atul Ramachandran, Filed Inc
Design vertical AI around delegating and supervising complete workflows so customers can gain time back.
Shipping Products When You Don't Know What they Can Do — Ben Stein, Teammates
PM for autonomous-agent products requires shipping despite genuine uncertainty about what your own product can do — a new discipline is emerging.
Don't just slap on a chatbot: building AI that works before you ask
Stop bolting chat onto products — proactive, context-aware AI that acts inside the existing workflow beats reactive chatbots.
Personality Driven Development: Exploring the Frontier of Agents with Attitude
Giving agents explicit forms and personalities is great branding and onboarding shorthand but inherits decades of human anthropomorphic expectations you must manage.
OpenClaw in Your Hand: Building a Physical AI Terminal - Lech Kalinowski, Callstack
A useful physical AI terminal can keep the device simple by separating responsive local interaction from heavyweight backend intelligence.
Build for the Memo, Not the Demo — Shawn Chan, China Resources Holdings
Financial AI earns trust through traceable evidence, consistent numbers, and visible uncertainty.
Shipping something to someone always wins — Kenneth Auchenberg (ex. Stripe, VSCode)
Build for sub-day OODA feedback loops with named real users — continuously viable products beat big-bang launches, especially in AI where iteration speed is the moat.
Designing AI To Scale Human Thought — Jun Yu Tan, Tusk
AI products that augment human thinking (blind-spot detection, cognitive partnership, proactive guidance) beat those that automate the human out of the loop.
The Next Game Engine Won't Have a Manual — Arturo Nunez, Nereu
An AI-native game engine can make creation more accessible by mapping player vocabulary onto reusable engine behaviors.
Your AI Agent Isn't an Engineer: The Art of Thoughtful Anthropomorphism
Stop marketing AI agents as engineers — frame them as augmentation tools with transparent capabilities and limitations to build durable developer trust.
Agent Output Is Not UX: Rendering Layer Your LLM Pipeline Is Missing - Bala Ramdoss, Amazon Lens
Agent products need a version-aware rendering layer that turns model results into safe, responsive actions users can complete.
Invisible Users, Invisible Interfaces: Accelerating Design Iteration with AI Simulation - Alex Liss
Simulated user personas can give designers an inner feedback loop that surfaces friction across a category at scale before any human user study runs.
Build Dynamic Products, and Stop the AI Sideshow — Eliza Cabrera (Workday) + Jeremy Silva (Freeplay)
Differentiated AI products integrate AI directly into core product strategy via a crawl-walk-run progression, not via quarantined 'AI features' that fail to solve real customer pain.