🎙️ How I AI: Grok vs. Dot vs. Muse vs. Duet + How Figma’s PMs actually use AI every day

The Era of Tool Fatigue

We are currently drowning in a sea of wrappers. Every week, a new founder pitches me on a personal assistant that is supposed to change my life. They all promise the same thing: it will understand you, it will manage your calendar, and it will magically surface the information you forgot three months ago. But if you look at how actual product teams—like the folks over at Figma—are using these tools, the reality is much more fragmented and practical.

As builders, we often fall into the trap of looking for the one platform to rule them all. We want the single interface that solves the workflow. But the current landscape of Grok, Dot, Muse, and Duet shows us that we aren't heading toward a monopoly. We are heading toward a specialized toolkit where the winner isn't the smartest model, but the one that fits into the existing friction of a workday.

The Grok vs. Dot Dilemma

Grok is an interesting beast because it thrives on real-time noise. For a builder, it serves as a pulse check on the immediate zeitgeist. If you’re trying to figure out if a library is breaking or what the sentiment is on a new release, Grok provides a layer of social context that ChatGPT lacks. However, it’s noisy. It’s the tool you use when you want to know what’s happening now, even if the accuracy is sometimes questionable.

On the other hand, you have Dot. Dot is designed to be the empathetic listener—the memory bank. For a founder, the value proposition here is simple: recall. We spend half our lives looking for that one link or that one decision made in a meeting three weeks ago. Dot tries to solve the context window of human memory. It’s less about generating code and more about maintaining the thread of a project.

Muse and the Search for Meaning

Muse represents the shift toward deep thought. When I look at how PMs are utilizing these tools, Muse often comes up in the context of synthesis. We are generating more data than ever before. We have transcripts, Slack logs, and Jira tickets. Muse isn't just a chatbot; it’s a canvas. It’s for the builder who needs to map out a complex architecture or a product roadmap without getting lost in a linear chat interface.

The common thread here is that none of these tools are replacing the human at the center. They are acting as cognitive peripherals. If you’re building in this space, stop trying to build a better chatbot. Start building a better way to handle the specific type of information your user is already struggling with.

How Figma Really Uses AI

Figma is often held up as the gold standard for product-led growth, so watching their internal teams use AI is a reality check for the rest of us. They aren't just letting AI write the product requirements documents (PRDs). Instead, they are using it to kill the blank page problem.

• Summarization: Taking a 45-minute sync and turning it into three actionable bullets.
• Prototyping: Using AI to generate dummy content or basic layouts so they can get to the feedback loop faster.
• Querying Internal Docs: Instead of asking a teammate where the spec is, they ask the bot.

The takeaway for founders is that the most successful AI implementations in high-performing teams are invisible. They don't feel like a new task; they feel like a shortcut to an existing one. If your AI tool requires a 20-minute onboarding session, you've probably already lost the battle for the user's attention.

The Duet and the Enterprise Reality

Google’s Duet represents the enterprise side of the coin. It’s boring, but it’s integrated. For a builder, this is the cautionary tale. You can have the coolest, most innovative standalone app, but if the big players can bake a 70% version of your feature into the suite the team is already paying for, you’re in trouble. Integration is the moat.

We are seeing a trend where 'good enough' AI that lives inside your email or your document editor beats 'great' AI that requires a new tab. This is why Figma’s internal use cases are so telling. They use what’s available where they are already working.

What This Means for Builders

If you are building an AI-first product right now, you need to be honest about which bucket you fall into. Are you a utility like Grok, a memory bank like Dot, or a workspace like Muse? Trying to be all three is a recipe for a bloated product that no one actually likes using.

The 'Founder Perspective' here is simple: stop obsessing over the model. GPT-4, Claude, and Llama are commodities now. The real value is in the plumbing—how the data gets in, how the context is stored, and how easily the user can get the result out and into their actual work.

The goal of AI isn't to create more content; it's to reduce the time between an idea and its execution.

We are entering a phase of AI skepticism where the 'wow' factor is gone. Users want to know if this tool will save them ten minutes today. If the answer is 'maybe, after you train it,' they won't use it. The Figma teams use AI because it helps them move fast. That’s the only metric that matters.

Final Thoughts for the Skeptical Founder

Don't get distracted by the feature wars between these platforms. Grok will always have the Twitter data. Google will always have the Workspace data. Your job as a builder is to find the data that they don't have—the specific, messy, internal context of a niche workflow—and build the bridge to it.

The future of work isn't a single AI that does everything. It’s a series of small, specialized agents that actually talk to each other. Until then, we’re all just toggling between tabs, trying to find the one tool that actually remembers what we said yesterday.

Read the original at Lenny's Newsletter →