Why aren't more people talking about this? It's almost impossible for non-technical founders to find tech partners in 2026 I went over to check out Upwork after 4 years. It hasn't changed one bit lol and most project posts get hundreds of bids. I ended up chatting with a few of these founders and they told me it's a massive pain - on platforms like Upwork or off them. - They don't know when they're being lied to - They don't have a nuanced understanding of tech to find red flags - Everybody looks identical, writes identical, sounds identical (thanks, AI) - They only REALLY know after doing a little work with them (but that puts all the risk on the founder - time + money) The alternative is AI. Some #founders go for vibecoding which itself is a trap. (I'll rant about that another time) So what the hell do non-technical founders do? Honestly, I want to help them.

sometimes Claude is a tool

Claude is down Codex is down And suddenly there are no new posts on LinkedIn 🤔

Oct '25: founder has app idea. Nov '25: I go to NYC & play pickleball (for the first time) with him. Aug '26: it's live on the App Store! It was a community app idea, so I had to meet the community. I flew to the US in Nov 2025. And they were awesome! Each member was so different, and yet pickleball is what brought them together and made them a family. Bharat is that founder. He knew everyone, and everyone knew him. He really cared about the people and the game - and wanted to build something to organize the community. That's when I knew I had to work with him and his team. And that's also why it's so great to see the app go live! pickleheart.org

Feels like n8n / zapier / ifttt and the likes are done? At least the way we've known them. I already saw Relay go under just this month. The founder just went to Google with no explanation, and they sunsetted a seemingly good business 🤷‍♂️ It sucks - because I used to use Relay, and I was also all in on n8n until a year back. They were great for piping data between other tools. I built internal processes and client solutions on them for a while. But then things changed - I moved to building my own tool connectors (custom n8n alternatives) - I further moved to building my own tools entirely that directly integrate with each other So now I don't feel the need for such connectors / piping at all. To me, that entire use case has been abstracted out. What creeps me out is how recently it was actually at its peak, and now it's not even something I think about. My recommendation to business owners running n8n workflows internally is to move over to custom built alternatives and Claude/OpenAI functionalities. It's orders of magnitude better for internal processes. Speaking from experience.

"How do I ensure my AI system's results good?" - I'm building self-eval into an AI product rn and you WON'T like the answer I found The product I'm building generates previews of medical procedures before they actually happen. For the longest time, we just assumed the results were good, because they looked good to my untrained eye. Until someone looked at it and said - "your results look unrealistically good - there's overcorrection here" Yikes. OK so then we grabbed a bunch of actual before/after images to benchmark against and she was right. The AI had to be tweaked. But tweaked to what exactly? To find out, we built a self-evaluation benchmark Here's how it works: - Upload REAL before+after pairs - AI generates its own 'after' image without looking at real 'after' - Another self-eval AI compares the AI generated v/s real 'after' images - It scores & suggests changes to the AI system That works well. BUT the evaluator is also AI 🫠 So how do you know if the evaluator itself is giving us good results? You evaluate the evaluator. And recursively keep doing more benchmarking / eval steps, until a human (me in this case) can check it manually 🤷‍♂️ It does comes down to 'vibe-QA'-ing eventually. So here's my answer to 'how to ensure my AI system has good results': You can't know for sure. All you can do is keep auto-evaluating, until someone you trust can take a look and say it's good If you have a better answer, I'm all ears.

xitter.cf/levelsio/status/209375… ok the age of interdimensional cable has begun

Uncertainty sucks. 6 clients right now, but dunno if I'll have any in November. So I'm trying to brute-force some certainty into my life. Most service businesses work like this. We have a ton of leads one week, and then nothing for a month. And we're just hoping for something to land before our cashflow dries up. I thought that was the nature of the business. But I'm sick of it! So I'm trying all kinds of things. This one just worked: - Find person / project post that meets my parameters (on project platforms, forums, etc) - Connect with them there - Also find them & connect on social media / email - Have good portfolio and credibility - They often respond in one of the places - Schedule a call 2 discovery calls booked after a week of this. Only works with good portfolio + relevant credibility / proof. Grateful for past me busting his ass 🫡 And the experiments continue.

I'm cancelling most of my software subscriptions. But intentionally letting some of them stay I've been moving to self-built versions of the SaaS tools I used. No monthly costs, I get to customize them and there's no duct-taping between them. This needs to be done thoughtfully tho. Here's what I did rebuild internally so far: - Call recording / transcription tool (removed Fireflies / Otter / Granola) - Task management (removed Monday / Notion / Trello) - Small scale transactional and marketing email (removed Resend / Brevo) - Internal dev processes w/ Claude+Vercel (removed Lovable / Replit) - Building a work messaging platform (will remove Slack) - Figma - replaced, not rebuilt. Using Google Stitch and Claude Code for HTML prototypes instead of Figma, unless explicitly necessary What I am intentionally keeping in place: - Zoho billing (I don't want to mess with compliance stuff yet) - Infra - DB, Hosting, Domain services - Linkedin Premium (can't rebuild that - yet lol) - Frontier AI providers - OpenRouter / Claude Max Cleaning up feels good. And owning your own stack, built custom for your own business feels great.

Why don't I ever hear of anyone using @Netlify? it seems legit

I ran an email campaign on 5389 users. $0 spent, no external services. 100% Gmail + Sheets. It took me 30mins to set up the system and test it. It took 5 days to send all the emails. Gmail on G-workspace has a hard cap of 1500 emails a day. So I spaced it out. But that's more than good enough for a small business. Google appscript is weirdly underrated! Pro tip: Set up the email sender on a different domain if you want to do this often. If you want my sheet with the auto-sender, just ask :)

i just listed blurstudio.ai on outliked, the leaderboard where you participate by posting and win based on likes. every like on this tweet pushes it closer to #1. do your thing 💗 outlike.lol/?via=blurstudio.…

This is one of my 'salesman' shirts. I HATE my salesman shirts. But I'll be wearing one of these everyday until I build steady leadflow. It's a Saturday. I usually wear something way more chill on the weekend. Like a tee. But I'm using this as psychological warfare on myself, to do stuff I'd have happily avoided given half a chance. That stuff is sales, outreach, finding leads and getting creative with it. So if you see me wearing something boring like this. It means my lead funnel is still under construction. I have 2 offers from @blurstudioai that I'm promoting: 1. If you know someone struggling on a Bubble app, I'd love to help them decide whether now is a good time to move off. 2. If you're running a Bubble agency or are a freelancer, I'd love to help you extend your relationship with any client that's outgrown Bubble by partnering

New iOS app submitted today. Funny story - we started it on Bubble and finished with custom code In Oct '25 I got in touch with this founder. They were keen to build with Bubble native mobile and implement in-app purchases. In Dec '25 I brought in the amazing Chris Schrade to co-build it with on Bubble. In Feb '26 Bubble released support for In-App Purchases. And we started building. In April '26 we started testing. We found some strange bugs and spent a few weeks trying to debug. At the end of April, we finally reached out to Bubble support. In June '26 Bubble actually responded. By that time the founder was quite frustrated with the wait. I couldn't keep Chris hanging on either. In July '26 we decided to rebuild the app with React Native. That's also when @BlurStudioAI went from agency to advisory. In early Aug '26 I finished the build and started testing. And today we submitted the app. In Sept '26 the promotion and marketing campaigns kick off. And the business actually launches. At the end, the founder gets set up with a solid code setup where they can make incremental updates with AI, manage branches and make sure the app is functioning well - without depending on me or any platform. No better time than right now to turn ideas into products.

Not using any vibecode tools anymore. There's too little control and things go sideways all the time. If anyone is deciding what stack to use for a really good app, this how I do things for my clients: - human for architecture - human for setting up DB, infra, migrations, etc - AI-assisted coding (used like a junior dev) - human for supervising the code writing - AI-assisted code testing - AI-assisted e2e testing (first pass) - Human for final e2e testing The thing that takes the most time has always been the writing of the code itself. Typing is just slow, even if you know exactly what you're going for. I've always thought of app logic as pseudo code - and then written the lines to fill out that vision of the logic. Vibecoding takes too many architecture decisions based on bad assumptions. And any gaps in your requirements compound.

I DID IT!! All emails with this client have been on a single thread. And I just sent the 100th email on that thread There's got to be an award for this

I'm lowkey scared of outreach and sales. Dunno why. But I've decided to become a pro at it this year. It's not going to be easy obviously. But it's necessary. I'm just going to do it until I get really good. I tried to throw AI at the problem and ignore the very thing I've been telling people - "don't automate until you understand how it works". No. I need to practice what I preach. So if you could give me one tip for great B2B outreach, (which I would really appreciate 🥹) what would it be?

xitter.cf/forgebitz/status/20897… So the best places to have content for AI to pick up is basically great articles on your own website or guest blogs? Looks like SEO is the new GEO (Gemini cited sources from Reddit ironically lol)

I acquired a little SaaS called @atomicfusionio in 2022. In Sept '26, I'm winding it down and taking it in a different direction. SaaS has changed. Nocode has changed. The world has massively evolved in the last 12 months. Building, running and growing Atomic Fusion has been one of my best experiences so far. The 7k users, the Bubble community, and everyone I met and got to know in the process have helped me grow a lot. Thank you so much! One thing is unchanged: I love helping entrepreneurs build great businesses, with products and AI deployments. And that will continue. I'm just not sure exactly how. So I need your ideas: > What do you think Atomic Fusion should pivot to?

How I avoided outreach work this weekend: Notetaking suchs. So I took matters in my own hands. This bad boy detects when I open a meeting, prompts me to record/transcribe, plugs the transcription into my management system (also custom) and I never have to open a notetaker app (or pay for one) again, and my notes live on my device.