Notes on mx.works | AI Studio Building Apps & Automation Tools

You vibecoded a demo, not a product

Thu, 13 Aug 2026 00:00:00 +0000

We vibecode too. Plenty of what we build starts as something generated in an afternoon, and that isn’t a confession, it’s just faster than sketching. The demo works, it looks like the thing, and for a day it feels like you skipped six weeks.

Then you try to put it in front of people who don’t know where not to click.

A vibecoded prototype is a demo of the happy path

That’s the whole definition. A vibecoded prototype is software that proves the path where everything goes right: someone signs up with a real email, uploads the file you had in mind, pays with a card that works, and the screen does what the screenshot promised.

Nobody's asking for the chat window

Wed, 29 Jul 2026 00:00:00 +0000

Somewhere in the last two years, adding AI to a product turned into adding a chat window to a product. A little bubble in the bottom right, a sparkle icon in the toolbar, a panel that slides out and asks how it can help. It became the default move, and most of the time nobody asked for it.

We went through this ourselves. When we rebuilt tidi earlier this year, the obvious version had a chat panel on the side. Every canvas tool has one now. We built the product without it, and the AI reaches the canvas through MCP instead, which means an agent can create and move and read cards without a human typing into a box inside our app. The interesting part is what that does to the interface. There’s nothing to explain, no empty state telling you what to ask, no second way to do everything you can already do.

We rebuilt tidi from scratch

Thu, 04 Jun 2026 00:00:00 +0000

tidi started as a visual bookmarking idea, drifted into agent infrastructure, and stopped feeling like anything we wanted to use. So in May we threw the whole direction away and rebuilt it from scratch.

The first version chased a big abstract idea. Agent memory, company knowledge, infrastructure for other people’s workflows. On paper it was a market. In practice we never opened it. That’s the tell. When you stop using your own product, the product is wrong, not you.

Agents are users now

Wed, 13 May 2026 00:00:00 +0000

An agent interface is any way an AI agent can use your product without a human clicking through your UI. That includes MCP servers, command-line tools, and clean public APIs.

Here’s the part most founders miss. When you ship one of these, your product becomes something an AI can recommend, invoke, and use on a buyer’s behalf. No homepage visit needed. No ad spend. The agent is now in the room with your customer, and it picks the tool that’s reachable. That’s a brand new distribution channel, and it’s mostly empty.

ClawDeck is shutting down. The open source stays.

Wed, 06 May 2026 00:00:00 +0000

I built ClawDeck in a weekend in February. An open-source kanban board where your AI agent picks up tasks, moves cards, and works through your queue on its own. I figured a handful of people would use it.

486 users, 3,800 tasks, and almost 200 GitHub stars later, I’m shutting it down.

Not because it failed. Because it worked well enough to show me what I actually needed to build.

Small teams got dangerous

Sun, 26 Apr 2026 00:00:00 +0000

Obsidian has 7 full-time employees. Around $25M ARR. A $350M valuation. No VC, no scheduled meetings, no stand-ups. About a million users per person on staff. And they built all of it before AI was even a serious part of the conversation. Just the right team, the right product, and a flat structure by design. That used to be an anomaly.

It’s becoming the template.

Greg Isenberg put it cleanly: “Obsidian is a $350M company for a note taking app built by 3 engineers working remotely. No other time in history was something like this possible.” He’s right. Seven people total, three of them engineers. The rest of the headcount is design, support, and product. No layers. No middle. Deep focus, as Steph Ango (@kepano) describes it, is the actual operating principle.

Figma is a sketchpad now

Thu, 09 Apr 2026 00:00:00 +0000

I have been a product designer for twenty years. UI, UX, branding, everything in between. I have spent thousands of hours in Photoshop, Sketch, InVision, Figma. I have shipped a lot of interfaces. I love design. It is most of who I am professionally.

The way I work has completely changed in the last year, and I want to talk about it honestly, because I am seeing other designers and teams go through the same shift.

Taste is still doing the work

Fri, 27 Mar 2026 00:00:00 +0000

I wrote something on andresmax.com today about taste being the new competitive moat. The argument came from watching six different Lovable prototypes that technically worked but felt like nobody made them. The features were there. The judgment wasn’t.

That’s the AM take. The builder take is more specific.

Last week I did a full pass on this site. Accessibility audit. Bolder typography. Scroll reveals. Parallax on the hero. Rebuilt the contact page. None of those are big features. Each one is a small decision. But stacked together, they’re the difference between a site that looks like something and a site that looks like it was generated.

Introducing lst.so: the missing piece was a shared task list with AI

Thu, 19 Mar 2026 00:00:00 +0000

I was losing tasks across five products, three AI agents, and my own head. Something had to give.

I had already built ClawDeck, an open-source kanban board for AI agents. It worked, people liked it, agents used it daily. But the more I used it myself, the more I realized a kanban board wasn’t what I needed. I needed a list. Not columns and cards and drag-and-drop. Just a flat, minimal task list that I could share with any agent, not just OpenClaw.

Claude Code + Obsidian changed how I work

Tue, 10 Mar 2026 00:00:00 +0000

I’m building, shipping, and running five products and a staff augmentation company. At any given moment there are 15 things competing for attention and I’ll have three new ideas before lunch.

For years my system was: write it down somewhere, forget where, rediscover it two weeks later. Obsidian helped. But Obsidian alone doesn’t do anything with the notes. It’s a second brain that doesn’t act.

Then I saw Greg Isenberg’s tweet about using Obsidian + Claude Code as a “24/7 personal operating system.” The idea clicked immediately. Not because it was new to me, I was already using both. But because I hadn’t connected them. My vault was over here, Claude Code was over there, and I was the bottleneck in between.

ClawDeck is taking off

Fri, 20 Feb 2026 00:00:00 +0000

Update, May 2026: ClawDeck’s hosted version is shut down. The roadmap further down never shipped. The MIT-licensed open source stays up, and the full story of why it ended is in ClawDeck is shutting down. The open source stays. The post below is left as it was written.

A few weeks ago I shipped ClawDeck as a weekend project. A kanban board where your OpenClaw AI agent can pick up tasks, move cards, and work through your boards on its own.

We built nod.so to save us $3k in bills

Thu, 12 Feb 2026 00:00:00 +0000

I was paying around $250/month for email services I barely used.

Kit for one product. Mailchimp for another. Separate accounts for Gratu, tini.bio, mx.works, ClawDeck, andresmax.com. Five products, four paid plans, four logins. And I was using maybe 5% of what any of them offered. No drip campaigns. No automations. No landing pages. Just subscriber storage.

So I built my own.

nod.so

nod.so is a subscriber management tool I built in 48 hours using Claude Code. It’s a Rails 8 app running on Render with Resend for transactional emails. It currently manages over 2,770 subscribers across 7 lists, all from a single dashboard.

Building ClawDeck

Tue, 03 Feb 2026 00:00:00 +0000

Over the weekend I built ClawDeck, with the assistance of my bot, Maxie. It’s a kanban mission control for your OpenClaw bot.

The idea is simple. Take the simplicity of a kanban board and put the full power of an AI agent behind it. You can organize your projects, to-do lists, whatever you need. But the real trick is that your bot can interact with the boards directly.

Gratu goes to TikTok

Fri, 23 Jan 2026 00:00:00 +0000

This week I started posting TikTok slideshows for Gratu.

The format is simple: 5-slide photo carousels. Emotional hook on slide 1, story in the middle, app screenshots and CTA at the end. I’m posting 2-3 times per day on both TikTok and Instagram Reels.

Two content types so far. Story slideshows that promote the app directly, and value listicles that grow followers. The direct promos drive downloads. The listicles build an audience that might convert later.

Shipped tini.bio 2.0

Mon, 12 Jan 2026 00:00:00 +0000

We shipped tini.bio 2.0 this week. Complete rebuild from the ground up.

The old version worked. But watching how people used it, I realized the CMS-style interface was too complicated. Separate sections for projects, posts, links. Too rigid. Too much like building a blog when all you wanted was a simple, beautiful page.

So we threw it all out.

What’s new

The biggest change is the editor. Everything now lives in one place. Your bio, links, images, content, all built using blocks you can drag, drop, and arrange into any layout. No more navigating between sections. No more wondering where things will appear.

Small businesses showed up

Mon, 29 Dec 2025 00:00:00 +0000

Something shifted with tini.bio over the past few weeks. Signups are up around 150% from normal. About 20 new users a day now, which for a quiet little product is a lot.

The numbers are nice. But the interesting part is who’s signing up.

I built tini.bio for creators and freelancers. A simple page to hold your links, show your work, look professional without fussing with a website. The classic link-in-bio use case.

Just Build the Thing

Mon, 15 Dec 2025 00:00:00 +0000

I keep seeing the same pattern play out in the startup ecosystem.

Founders spend two months in “discovery.” Workshops. User interviews. Competitor analysis. Stakeholder alignment sessions. A beautifully formatted research deck. Wireframes for every possible flow.

Then they run out of budget before building anything.

Discovery is Expensive Guesswork

Here’s the thing no one wants to admit: most discovery work is prediction dressed up as insight.

You can interview users, but they’ll tell you what they think they want, not what they’ll actually use. You can analyze competitors, but that tells you what worked for someone else, not what will work for you. You can workshop until the whiteboard runs dry, but you’re still guessing.

AI flipped my design workflow

Sat, 06 Dec 2025 00:00:00 +0000

I used to spend days in Figma. Pixel-perfect prototypes. Auto-layout nightmares. Obsessing over spacing and hover states before I even knew if the idea worked.

The process felt productive. Dragging rectangles, tweaking corner radii, duplicating frames for every possible state. But looking back, I was doing the same work twice. Once in the design tool, once in code.

Now? I sketch rough flows on my iPad when I need to think through something complex. Quick boxes, arrows, annotations. Nothing pretty. Just enough to externalize the idea and see if the logic holds. Sometimes I skip this entirely and just describe what I’m building.

Hello World

Fri, 28 Nov 2025 00:00:00 +0000

For those of you who know me, you know I’ve been building products and teams for over 18 years. Design, engineering, strategy. Startups, agencies, big companies. I’ve loved every minute of it.

But I’ve always built for others.

Today that changes. I’m announcing mx.works: an AI-first product studio.

What This Means

mx.works is where I ship my own product ideas. Not client work. Not consulting. Products I own, from concept to launch to growth.

Building This Site With AI

Sat, 25 Oct 2025 00:00:00 +0000

I just shipped this site. Built it myself, as one person, in about a week of focused work.

Here’s how.

The Stack

Figma for design. I started there because I think visually and I needed to figure out what this thing should feel like before writing any code.

Hugo for the static site. Fast, simple, no JavaScript framework bloat. Markdown for content, Go templates for layouts, deploy anywhere.

Tailwind CSS for styling. I know some people hate it. I love it. Keeps me in the HTML, moves fast, easy to maintain.

A Forming Idea

Sat, 18 Oct 2025 00:00:00 +0000

This idea has been forming for a while.

For years I’ve watched the same pattern play out. I help someone build a product. It ships. It grows. They own it, I move on. Repeat.

I got good at the work. But I never got any leverage from it. Every project started from zero. Every relationship ended when the contract did. I was building value, just not for myself.

The Question

At some point I started asking: what if I just… did this for myself?