The Fatigue of the All-In-One Doc

…and the inevitable pivot toward agentic platforms.

The UI is the Friction: Eviction of the Lego™ Apps is Intevitable

There is a specific kind of exhaustion reserved for the productivity enthusiast. It isn’t the tiredness of work done; it is the tiredness of work organized. I experience it often. I’m sure you know exactly what I’m talking about.

For the last decade, we have been sold a very specific, alluring promise: that we are all software developers in disguise. No-coders with vast energy and domain expertise, always ready to bring order where chaos reigned.

We were told that if we were just given enough “blocks”—tables, canvases, kanban boards, and buttons—we would build our own Nirvanas. We flocked to Coda. Others migrated to Notion. We spent our weekends gardening our wikis, pruning our databases, and aligning our properties with what was undeniably precisely what our businesses and organizations needed.

But the wind is changing. As I’ve written before regarding Coda, and as evidenced by recent rumblings in the Notion community (specifically a pointed observation by @lkr regarding the “heaviness” of maintaining these digital gardens), the “Maker” era of productivity is hitting a wall.

We are witnessing the death of the Canvas and the birth of the Agent. And frankly, I am fatigued. I’m ready for this. Perhaps it’s because I’m so damn old. I am that. But I’m also a rational realist. If there’s a way to work better and faster, with greater clarity and quality output, it’s difficult to reject. After all, isn’t this the mantra of Coda? It’s why we’re here, right?

The fallacy of the “Blank Canvas”

The premise of Notion and Coda was liberation. They freed us from the tyranny of rigid, single-purpose SaaS tools. Why use Asana, Google Docs, and Trello when you could mash them all into a single, elegant page?

The problem, as it turns out, is that “mashing” is manual labor.

When you buy a purpose-built tool, you are paying for an opinion. You are paying for someone else to have decided how a “Project” relates to a “Task.” When you open a blank Coda doc or a Notion page, you reject those SaaS-bound opinions. But you also absorb the responsibility. You must have that opinion. You have to be the Product Manager, the UI Designer, and the QA Engineer of your own workflow.

Initially, this felt like power. Now, it feels like overhead. And it’s a sizeable overhead.

The post from @lkr captures this sentiment perfectly.

It’s the realization that the tool has become a pet. You have to feed it, clean it, and groom it. If you step away for a week, you come back to entropy—broken views, outdated statuses, and a dashboard that demands input before it gives insight. Slowly, we are realizing that we don’t want to build the tool; we just want the result. We wanted results from SaaS tools, which gave us a different form of fatigue - workaround fatigue. That’s why we’re here.

The backlog of usability and functional UI requests, well documented in this community, serves as a graphic reflection of UI fatigue.

The Agentic Shift: Results over Rituals

This is where the trend line spikes. Slowly, we are moving away from User Interface (UI) and toward Agentic Interface (AI). It’s a snail’s pace today, but that will change. Most of us reject it on grounds that are occasionally dubious, but often rational and well-reasoned. Undeniably, in the not-too-distant future, we won’t build solutions with Lego™ blocks.

In the UI era (the Notion/Coda era), the user is the engine. You click the button. You drag the card. You filter the view. The software is passive; it waits for you to manipulate it.

In the Agentic era, the software is active.

Consider the difference between a dashboard and an analyst.

  • The Dashboard (Notion/Coda): Displays a graph of declining sales. It waits for you to see it, panic, click into the data, analyze the root cause, and draft an email to the sales team.
  • The Agent: Notices sales are declining. Analyzes the root cause (seasonal variance). Drafts the email to the sales team. Pings you on Slack: “Sales are down 5% due to the holiday; I’ve drafted a memo to the team to reassure them. Approve to send?”

The former requires you to “live” in the app. The latter allows you to live your life, which is tantamount to doing your job better, faster, and more successfully.

The increasing trend of “Headless” Productivity

This is why we are seeing a shift away from the beautiful, all-encompassing UI. The aesthetic of 2020 was a perfectly curated Coda workspace with aesthetic headers and perfectly tag-matched databases.

The aesthetic of 2026 will be a text message.

The fatigue with Coda’s complexity was the canary in the coal mine. We are seeing this pattern accelerate. Users realized that to get the most out of Coda, they essentially had to learn a proprietary programming language (Coda Formula Language). Now, Notion users are feeling the weight of their own creations.

We wittingly absorbed the responsibility with intention and delight.

The market, which is us, is correcting now, because it can. Agentic platforms continue to inch closer to the reality that they can reliably assemble the Lego™ blocks for us faster and with greater precision. But that’s not all! They can maintain our creations, patch them when features change, and fix them when workflows abruptly deviate. All without us asking.

We are seeing the rise of platforms that don’t ask you to design the house; they simply ask where and how you want to live.

This isn’t to say Notion and Coda will die. They will remain excellent tools for specific use cases—wikis, static documentation, and collaborative scratchpads. But as a proactive human-driven, tedious operating system for work? Probabalistically, extremely unlikely.

The future isn’t a better set of Legos™. It’s a hired contractor who builds the castle while you orchestrate the architecture of your professional work in the context of your enterprise goals. The “Maker” movement is being replaced by the “Manager” movement—where we manage AI agents, not database properties.

“Properties” will remain critical infrastructure elements. Secure, AI-competent elements capable of agentic interaction will thrive. Others will die slowly.

The user interface of the future isn’t a canvas. It’s a conversation.

4 Likes

I can relate to the idea, the burden of updating docs to remain relevant is tangible.

how much time and effort are needed to set up all these agents?

agents are instruction based and I wonder what it will take me to define my expectations and intuitions.

ideas?

cheers, christiaan

2 Likes

Agree. Any recommended resources to learn how to set-up agents @Bill_French ?

1 Like

This might be the only part of your post that I agree with. If the coda team kept up with non-AI related features (sharing pages, more secure ways locking filters, better display options) I would have zero complaints. I don’t like coda despite having to build & maintain, I like because I get to build & maintain. There’s a reason Lego are the most well known, long lasting, radical toys ever.

I also think you are really overselling the usefulness of AI.

1 Like

It’s pretty simple to get started.

Here is the revised “Simplest Steps”:

  1. Instrument: Point Antigravity (or Claude/Cursor) to the Coda MCP.
  2. Create (Prompt): Tell it to build the structure.
  • “Build a new Coda doc called ‘Project Alpha’ with a tasks table and 3 random tasks.”
  1. Interact (Prompt): Verify access.
  • “Check for open tasks.”
  • “Mark the first one as done.”
  1. Solidify (Workflow): Turn that interaction into a reusable tool.
  • “Save this logic as a new agent command called /status”.
  1. Scale: Run /status anytime.

This moves you from ChattingDoingAutomating in less than 5 minutes. No schemas, no “Laws” required upfront—those are just guardrails the agent handles silently.

… I wonder what it will take me to define my expectations and intuitions.

The “instruction burden” is the new friction, but it is fundamentally different from the “maintenance burden” we are leaving behind. It’s a bigger conversation, but here are some thoughts:

1. How much time/effort to set up? It is front-loaded, similar to onboarding a new junior employee. You don’t just “install” an agent; you have to orient it.

  • In the Maker Era (Coda/Notion): You spend hours building the “machinery”—the tables, formulas, layout, and buttons. If you stop maintaining it, the machine breaks (doc rot).
  • In the Agentic Era: You spend those hours defining the Standard Operating Procedure (SOP)—(@Max discusses this a lot), the prompts and context.
  • The Difference: Once an agent is “trained” and given access to your tools (via things like MCP), it doesn’t “rot” the same way. It adapts. The effort shifts from repetitive clicking to periodic reviewing.

2. Defining Intuition and Expectations This is the hardest part, but also where the technology is moving fastest.

  • Context replaces rigid instruction: “Intuition” is often just a high-fidelity understanding of context. Standard automation (Zapier) has zero intuition because it has zero context. Modern agents (LLMs) can read the entire history of a project, which allows them to infer “how we do things here” without you explicitly coding every rule.
  • Iterative Alignment: You won’t define your expectations perfectly on day one. You interact with the agent, correct it, and it “learns” (by you updating its instructions or context). It’s a conversation, not a compilation.

We are trading the fatigue of acts (clicking, dragging, updating) for the effort of intent (defining, reviewing, approving). I find the latter to be much more sustainable.

1 Like

I recommend nothing more than installing Antigravity - it’s free. Then complain like a banshee until @Bharat_Batra adds you to the MCP beta. Then, with MCP configured, as a completely untrained agent, modify one of your Coda documents or tables. The rest will be obvious - your mind will soon explode. :wink:

1 Like

I get that. It’s a great time to be alive, and I also like exploring Coda’s big basket of Lego blocks. But I’m also aware that there are times when rapid implementation and quick resolution trump playtime. There are also times when sustainable support and automation systems can be managed by machines. Less so for SMBs and consultants, but certainly for enterprises that thrive on efficiency and more sustainable outcomes that occur when I’m at the beach.

Indeed, but they haven’t. Workarounds seem to be required and, unfortunately, in greater numbers.

Perhaps. It’s a matter of experience. I have used Coda for a decade without agents and for only 6 months with agents. I’m roughly 1.7x more productive now than I was before. I hope to reach 2x by summer and maybe 3x by 2027.

2 Likes