What to learn, what to watch, what to skip. Focused on product design, AI-assisted workflows, and the emerging design-engineering overlap.
Figma Make now lets designers visually edit production codebases from within Figma. As of May 28, 2026, the limited beta supports direct property editing, annotation-based prompting, Git workflow, and pull request creation. The AI agent locates and modifies the relevant code when you adjust elements on canvas.
This is the design-engineering convergence point. For a Lead PD who codes, this is immediately relevant. Figma is betting the company on it: Make weekly active users grew 70%+ QoQ, and AI features drove Figma's Q1 2026 revenue to $333.4M (46% YoY growth). [1][3]
For SuperOps: If your design system uses Figma, Make + MCP integration means designers can push visual changes to production components without handoff delays. Worth prototyping with the Super Kit.
Andrej Karpathy coined "vibe coding" in Feb 2025. By May 2026, he's declared it dead — replaced by "agentic engineering": orchestrating AI agents against detailed specifications with human oversight. This is the methodology shift that matters.
SDD workflow: write a "constitution" (mission, tech stack, roadmap), then iterate through feature phases: plan → implement → validate → replan. Specs live in the repo as markdown. JetBrains published a DeepLearning.AI course on it. GitHub released Spec Kit as an open-source tool with 30+ agent integrations.
For indie iOS dev: This directly maps to how you should structure Time Left, WhatPay, etc. Write the spec once, iterate with agents, preserve decisions in the repo. [5]
Anthropic launched Claude Design in April 2026 — an experimental prompt-to-visual tool. Describe what you want in natural language; it generates presentations, layouts, and basic prototypes. Adapts to tone, style, and brand identity over time.
It's not replacing Figma for precision work. But it's expanding who can create: marketing teams, startups, non-designers. For SuperOps, this could accelerate internal deck/asset creation without pulling PD time. The real play is the API underneath — if Claude Design's generation model becomes an MCP tool, you could automate design system documentation or onboarding materials. [2]
Figma enforced AI credit limits in March 2026. The result: customers opted to buy more after hitting caps. Revenue outlook jumped $55M to $1.42B-$1.43B for FY2026. Stock hit 7-week high.
This is the template for how AI tools get monetized. For anyone building AI-assisted products: the credit/usage-based model works. Users pay when they see value. Worth studying the unit economics before launching any AI features in your own apps. [3]
Figma acquired AI-native creative platform Weavy, rebranded as Figma Weave. Adds generative image, video, and motion capabilities to the Figma ecosystem. 13M monthly active users now have AI generation baked into their design tool.
For product design: this eliminates the "go to Midjourney/Runway and come back" loop. Generated assets stay in the Figma context. Early days, but directionally significant. [1]
The role is shifting. "Design engineer" used to mean someone who bridges Figma and code. Now it's becoming "agent orchestrator" — someone who writes specs, directs AI agents, reviews output, and makes architectural decisions.
Key competencies emerging:
For Prabha: You already do most of this. The gap is MCP — Figma Make now supports custom MCP connectors. Worth learning to connect Super Kit to agent workflows. [1][4]
Real: Figma Make + visual code editing (46% revenue growth proves adoption). Spec-driven development (Karpathy's own pivot). AI credit monetization (Figma's revenue jump).
Hype: Claude Design replacing Figma (early, experimental, less precise). "English as the primary programming language" (interesting but premature). Generic AI design tools without domain context.
Skip for now: Generic "AI can design anything" claims. Most useful AI design tools are domain-specific (ITSM for SuperOps, fitness for your side projects).