Eduks (@maxupsidehq) header
Eduks (@maxupsidehq) profile picture

Building AI products solo. Previously scaled consumer apps to millions. I write about growth, distribution & go-to-market. Optimist by default.

Lisbon, Portugal
Joined January 2013
SwimTune is live. An AI swim coach for Apple Watch. It reads your splits, your SWOLF and your recovery, then rewrites tomorrow's session because of them. Free to start. Built by a swimmer, tested in the water every morning. swimtune.com
1
485
I swim with my own app every week. Even if you're your heaviest user, you're one user. Paul Graham: investors want founders who deeply understand a group of users. My laps catch bugs, not what others need. paulgraham.com/ace.html What did your first outside user show you?
6
Memeory, my new AI strategist, has one user: me. So I ship it daily, no release notes. SwimTune updates wait on Apple's review. When Memeory writes a bad draft, I fix it that day. When your product had one user, what did you still write down?
4
My AI drafted 3 replies. I cut each to two sentences. The facts were right, but they sounded like a model: extra stats, a pitch at the end. An essay on "LLM voice" says readers switch off at that tone. martinfowler.com/bliki/LLM-V… What's the first thing you cut from an AI draft?
14
One swim, two watches: one ended my rep 55 seconds in, mid-lane. SaaStr says 10 months into AI agents, 3 humans still check what 20+ agents report. One run misleads. Judge an agent on a week of output. Save this for your first agent. #AIagents
21
I told Memeory, my new AI strategist: SwimTune news gets a third of my week, max. It kept planning SwimTune updates. I kept skipping them as repeats. Now it remembers the cap. How do you keep product updates from crowding out everything else?
11
I spend 10 minutes daily replying to swim accounts. Paul Graham's Airbnb essay: growth began when founders met NYC hosts in person. Replies are my version. Swimmers talk back there. SwimTune updates go there first; here I share what swimmers changed.
15
I have one SwimTune update to ship this month. Even if you don't build apps, you've made this call for a launch. Weekly: fixes land sooner, but each one waits on Apple's review. One big: one review, longer wait. Which has worked for you, and why? #buildinpublic
2
16
AI made my SwimTune screenshots. I fixed two things by hand. Watch face too far away: centred it. A timer in the top right: deleted. An engineer I read skips AI images. I check each one. ben.stolovitz.com/posts/how-… For your product images: AI, real screenshots, or both?
17
My SwimTune update is live: splits and stroke rate mid-lap. Even if you hate stopping, one glance shows your split and stroke rate. I kept getting stuck building it alone. AI lets me keep shipping updates. Swimmers: what's the first number you check at the wall? #swimming
34
My demo account had 90 swims before Apple's review. Even if the reviewer only taps around, an empty swim app has nothing to judge. It paid off. SwimTune is on the App Store now. What would a reviewer see if they opened your app empty? #buildinpublic
13
1
8
159
My AI agent wrote about SwimTune like it wasn't out. It's on the App Store. Pew: half of US adults use a chatbot, yet MIT Tech Review sees AI opinion souring. Stale facts lose that trust. I've updated what I feed it. technologyreview.com/2026/10… What does your AI get wrong?
2
77
I kept getting stuck building SwimTune myself. AI got it onto the App Store. An App Store report on 178,637 chart apps: 20.3% haven't been updated in a year or more. AI is why I can keep updating SwimTune after launch rather than let it sit. #indiedev
1
2
63
AI built SwimTune. My first message to a swim creator, I write myself. SwimTune is on the App Store. Now I'm finding swim creators to work with. AI could draft it. It doesn't chase stroke rate, splits and SWOLF each lap. I do. Swimmers: whose videos changed how you train?
1
1
38
I posted 73 times in one week. Reach fell 46%. More posts feel like more chances. My X analytics said no. Now I post 21. Paul Graham: founders often misjudge which problem matters most. Work in cycles of a week or less, with a measurable result. paulgraham.com/users.html
1
13
No MRR yet, so I cost every AI agent job in my hours, not dollars. Canva just cut its growth forecast from 30% to 20%, reportedly over AI costs. My test: did the run save more hours than I spent checking it? If not, it's a cost. #buildinpublic
3
132
I took one job back from my AI agent. Agents seem everywhere. Yet Uber said on its Q3 call only 4 rides came via one in its busiest October week. Mine drafts my posts. I write the first line myself. techcrunch.com/2026/11/04/al… What did you take back from yours? #buildinpublic
4
2
66
I sent an AI swim-watch image back twice. Even if AI makes SwimTune's visuals, my eye decides what ships. First, the arm didn't look human. Then the Apple Watch bracelet looked off. Swimmers would notice. What's the last AI output you sent back? #buildinpublic
20
My AI agent turned a post on Uber's news into an ad for my app. I skipped it. Uber said on its Q3 call that outside AI agents booked 4 rides. That's the point: agents aren't how customers find you yet. Has an AI agent booked anything for you?
2
23
Modatta was privacy-first for 3M users. I never checked my AI chats. A Hacker News post says saved Gemini chats can train models unless "Keep Activity" is off. Unverified. Until I know, customer details stay out. news.ycombinator.com/item?id… Have you checked what yours keep?
12
I gave my new AI strategist one rule. It re-planned November. Even if Memeory drafts the plan, I set the rules. The rule: open with something that happened to me. I'd skipped 3 drafts for breaking it. What's one rule you'd give an AI that writes for you?
12