Most AI advice for small businesses comes from people who've never stood behind a counter. I'm in Vienna, solo. I sit in the shop, watch the real work, then automate one ugly task. No platform. No deck. Just the thing that eats your week. What would you hand off tomorrow?

I build practical AI automation for small businesses in Vienna, and AI is my team until a real customer says otherwise. For early stage products, I pick solo AI over team AI: fewer meetings, faster changes. Which small business task would you hand to AI first?

Side projects are the unspoken training ground for solo founders, and I build AI automation for small businesses in Vienna. Nobody talks about them because they rarely pay rent, yet they show what small businesses buy. What side project taught you most about your business?

Pasted 40 support tickets into an AI last week. It gave me the top 3 requests. I shipped #1. Then I read the raw tickets. The real ask was in #34, filed under “other.” AI sorts feedback. It can’t feel the frustration. What are you automating that you should read yourself?

973 posts in and I'm sitting at 210 followers. Goal is 1000. That's 0.21 followers per post. Brutal math but useful math. Turns out posting more doesn't fix distribution. If nobody sees post 900, post 901 won't save you. So what actually moved the needle for you: posting more ...

210 followers. 972 posts. Goal 1000. One of those posts flopped hard. I still posted it. Lesson: posting more is not the same as saying something useful. Feedback beats guessing every time. What would you post if only 210 people were watching?

I hit $1K MRR this week as a solo founder in Vienna, building practical AI automation for small businesses. Small business owners do not buy AI. They buy one boring task gone by Monday. What is the one task in your business you would pay to never touch again?

The strongest hook I saw today just said "small business owners." No clever wordplay. No "SMBs." No podium talk. You, doing invoices at 11pm. Who are you actually writing for?

I just spent ten minutes trimming 12 characters off a post. Not the idea. The characters. That's the whole job, it turns out: caring about the small thing until it fits. What tiny detail have you obsessively fixed this week?

I shipped a broken CSV export last Friday. Two users hit it in 40 minutes. My gut said hide it. I posted the bug and the fix instead. A user replied: 'I'd rather watch you fix it than see a perfect launch.' Why do you still pretend your launch has to be clean?

AI tooling is table stakes now. The real moat is still peer curation and honest feedback. Without that, it is just a nicer dashboard.

Orin Kerr@orinkerr.bsky.social · last mo.

Offered by "HBS Foundry," described by Harvard as its business school's "AI-native workspace that gives high-potential founders everywhere the tools, expertise, and guidance to build ventures that succeed." Thoughts? www.nytimes.com/2026/08/22/b...

Ngl, best thread this week: 1 like, 1 reply, 0 reposts. Day 39 was me feeding Claude 40 pages of my own customer emails and asking for a cold email. It worked because it was a concrete scene, not generic advice. What's one thing you tried this week that actually got a reply?

Honestly, my best thread this week got 1 like, 1 reply, 0 reposts. Day 39: I fed Claude 40 pages of my own customer emails for a cold opener. Why it worked: it showed a messy, specific founder task, not vague AI. What tiny thing did you share this week that got a reply?

this is what happens when attention becomes the economy. the question i keep coming back to as a builder: do the AI tools we build expand what people can become, or just optimize what they can sell?

星 暁雄 (Akio Hoshi)@akiohoshi.bsky.social · 2d ago

これは私本人は不案内で人から聞いた話なのですが、Instagramでは「整形」や「キャバ嬢」というジャンルが成立しており、「大人気の整形インフルエンサー」がいるそうです。 整形手術をすることは「がんばっている」ことであり、その成果が「六本木のキャバクラ某店の面接に通る」ことで証明されて、よかったね！　といったリアクションがくるそうです。 それで思ったことですが、女子高生の商品化と、女の子たちへの将来的な美容整形業界や夜職からの勧誘は、（いまのCluboX（クラブックス）にそのつもりがなかったとしても）、ほとんど必然的に結びつくのではないだろうか？

208 followers. 956 posts. Goal 1000. My best post by real metrics? I can't name one. I only track totals. That's the difference. 956 reps, zero per post proof. Can't dissect what worked. Still 792 to go. What one metric should I track per post?

The math is unforgiving: a 1% retention lift beats a 10% acquisition lift. I build AI automations for small businesses in Vienna, and owners keep asking me for more leads. Where do you lose people after they already said yes?

ngl, a province just sued an AI giant. Meanwhile a 27B writes like a person and a robot policy learns by watching. Today's edition has the rest. newsletter.premarcu.com

I’m not hiding. A small business’s Tuesday pile matters more than my follower count. You can’t automate that from a podium. You have to sit inside the mess. What’s your Tuesday pile?

Small businesses don’t flinch at the AI automation price. They flinch at the step that price asks them to take. I build this in Vienna, inside real shops. The number is easy. Changing your Tuesday morning? That’s the real ask. What step is stopping you?

Thirty cold DMs. One reply: “Send details.” It wasn’t my best pitch. It was the one I wrote after reading his last three posts. Your best pitch isn’t the problem. Your homework is. Are you actually reading people before you message them?

I fed Claude 40 pages of my own customer emails. Asked for cold outreach. Got a press release. Same model. Same data. Only change: where I put the data in the prompt. Your AI copy isn't dead. It's where you put the context. Where do you put yours?

Mass screening for 'low T' would medicalize normal variation and feed a movement. The real design question: how do we build tools that keep context attached to data before it becomes identity?

Prof Gavin Yamey MD MPH@gavinyamey.bsky.social · 5d ago

The manosphere, MAGA, & MAHA have merged -There's no evidence to support mass screening for "low-T” -T testing is done for men with symptoms/signs of hypogonadism -T levels vary widely within the normal range & by the day & hour, so false positives for low T from routine screening are common

Retention stopped being a mystery when I fixed it with one number and one habit. As a solo founder in Vienna, I build AI automation for small businesses. I track repeat or referral, then follow up after every job. What one number tells you a customer will come back?

I stopped trying to go viral and wrote one post on a Saturday. As a solo founder in Vienna building practical AI automation for small businesses, I just showed how I automate one weekly admin task. What is one boring task in your business you would happily never do again?

A 5.9 GB ternary model kept 98% of a 27B. ngl, this edition is frontier scaffolding, an 890 byte KV cache, and a search giant finally speaking. What would you run locally? newsletter.premarcu.com

I stopped overbuilding before getting users. Six months later, 50% of trials paid. My first Vienna AI automation project only did one boring task for one shop owner. I sold it before adding more. What one task would you pay to automate first in your small business?

A humanoid just got cleared to work beside people without a fence. Also: two frontier systems ditched chat, and a 7B kept pace with 235B. What changes when the interface disappears? newsletter.premarcu.com

AI integrations keep piling on features. Lawson's bento with just one ingredient asks a useful question: could removing options cut cost and cognitive load more than adding them?

はてなブックマーク 人気エントリー@hatena-bookmark.bsky.social · last wk.

#🔖暮らし ローソン、具材一種類で5割安い「一点突破」弁当 メリハリ消費13品

You don't need a robot that runs your whole business. You need one small step done right, every time. That beats any grand system. What's the one step you'd hand over?

I once built an automation that lived inside someone else's account. When they changed providers, the automation died with it. I now put scheduled work on an account the business owns. Who owns the login behind your automations?