What are you talking about? This is my substack and the article is a very in-depth write up that a lot of people found value in. Nothing low effort or purely promotional with this one.
It’s different in that it’s basically another abstraction layer. It has long-term memory (vs project by project) and constantly reviews its output to improve its skills and create new ones. Hermes can actually spin up codex or Claude code sessions for you as well. Telegram gateway is a huge plus as well. Codex and Claude mobile solutions right now are buggy and more for one specific thread of work vs a general delegator and chief of staff so to say.
Will be taking this for a spin imminently. What seems to be very unique here vs other canvas tools are the generative UI controls.
If you’re confused about why everyone is shipping their own canvas tool - this is what I wrote when I reviewed Stitch from Google in my newsletter:
“every SOTA (State-of-the-Art) model can already do this. Give it a prompt, it'll spit out an HTML design. Ask for 10 responsive variations, you'll get them. Stitch is a Ul and context harness on top of Gemini, in the same way Claude Code is a harness on top of Claude's models. This means every Al lab will likely ship their own version, and they'll all interoperate because at the end of the day it's all HTML and markdown.”
More generally, this is a competition of where the product development work starts and lives. The business value will accrue to those who become a destination, not a plugin.
AI can now power a new sort of tool that supports the entire process (not just coding or just design). So there’s no reason for Anthropic or any other lab to give that up to another tool.
Paper Design etc all have the same fundamental problem as Figma - they’re designed for a process that won’t exist as we know it within a few years (and arguably half way there already). All of these canvas-based design tools assume that people want / need to directly manipulate things on a canvas fist before building vs going straight to building it (and by building I also mean you can ask Claude to build you 10 prototypes for different directions so exploration is not dead, just starts elsewhere).
Whoever is able to act as the central store for all your context and be the integration point for all your agents running on top (including your proprietary ones) will have leverage in the ecosystem. Everyone else will be the supplier. The more context you have from different places (code, tickets, PRDs, markdown files, CRM, etc) and the deeper your agents plug into all of this, the harder it'll be to switch away to something else painlessly.
It is truly staggering to see a tool used by (almost) every large tech company in the world so quickly lose ground. I don't think they're done done, but their position in the value chain has shifted drastically.
Yep, my experience as well coming from Big Tech although they'll be buoyed by intertia for some time. And yes, source of truth just makes more sense in code, so I don't think there's much of a moat there either. I think there's still value in this collaboration layer, but collaboration will change substantially as well and take a different shape. The world where the design team spends weeks / months designing, refining, and doing crits on the same file just doesn't make sense now - it was driven by the high costs of implementation so the decisions had to be made upfront. That's going to change.
Long-term yes. Assuming we'll reach a world where an agent is your primary interface, has access to all you need, and will spin up dynamic UIs for you when needed.
I'm sure that's part of it but I think it's a very small part of the story. They've been pushing hard on internal AI creation tools for a while, and those clearly didn't take hold.
I don’t think so. Your product / tool has to be extremely specialized and deeply necessary to pull of a walled garden MCP. The more likely alternatives are that someone else comes along with a better (open) integration, the model internalizes your tool’s capabilities, or the AI labs themselves build 1st party competitors.
The result is less that I want to go to Figma directly and more that I just want to skip it entirely. So, assuming the power of these aggregator agents keeps growing, the onus is on these tools to create useful integrations or get subsumed by a model capability or another tool with a better integration. It sounds like your experience is similar - you bypassed the tools with bad integrations instead of going to them directly.
Definitely don’t believe AI is going to “one-shot” Figma. But Figma was built for a world where product design is a stand-alone step in a series of discrete steps with hand-offs between them. It was a place where design started and ended, for a while at least. If people don’t see it as a starting point anymore, it becomes a very different business & product.
I think that also explains why a bunch of companies are now all racing to build the same thing - the everything-in-one-place universal context store with their own agents on top of it. Linear, Notion, Salesforce, etc. Because the alternative is a much worse business to be in.
The combinatorial utility of different MCPs / APIs inside an agent is an interesting angle. Figma can technically plug into the same MCPs and use all the same models, but if the software design process doesn’t start in Figma anymore, it does not matter. The value will accrue to the point of integration (the agent).
Re: the value of good UI/UX
I think short-term, the value of good interfaces will actually increase - if anyone can easily build out the same product in 10 different ways, the best designed one that people actually want to use will likely be the choice. But that’s holding constant lots of things like distribution, type of SaaS, its place in the transaction stack, etc. So either guess would make a lot of assumptions.
It also appears (so far at least) that these models really struggle with front-end design. Something like /frontend-design skill is good but only gets you so far. It still requires a ton of steering to get it to a sensical place. So for now, whoever can steer it is still valuable. But I’m sure more and more of that will get codified and internalized by the model and the harness. So the design steering will become more and more abstract.
Long-term, we’re likely moving towards dynamically generated interfaces. Claude is already doing it with diagrams and charts in the chat. This opens up so many fascinating questions. What happens when UI doesn’t have to be one-size-fits-all, where each person may get their GUI generated with their preferences and context in mind? What happens to the design process when your UI doesn’t have to scale to a ton of user types and use cases? Will we even be designing UIs or something else entirely? Will Jakob’s Law still apply or will our individual GUIs diverge so much that I won’t be able to navigate your smartphone if I pick it up?