Aleksandr Protsiuk

@Finime2 @blaire_pang 12 apps in and the boring stack still wins. less time choosing tools, more time shipping

@a_protsyuk @blaire_pang Couldn't agree more. I've shipped 12 no-code apps this yea

startup founders, i’m trying to understand what you actually use for your backend what’s your stack? framework, db, auth, infra, deployment i feel like i see a lot of “ideal stacks” online but not what people actually use

@rivestack and they find out during a demo. that's usually what finally moves the needle on fixing infra

@a_protsyuk the localhost db part is the tell. that's not "shipping fine," that's one aws outage away from a really bad week. infra debt compounds way faster than feature debt

had a call with a founder yesterday. 18 months in, $400k raised, 3 devs. asked why they need a CTO. 'we're shipping fine'. opened their codebase. no tests. no migrations plan. payments on free tier. db on localhost in prod. they were shipping. just not the things that matter.

@rivestack velocity pressure is exactly when the worst code gets merged. sprint end rush is where most of the debt is born

@a_protsyuk the code review part gets skipped first when teams feel velocity pressure. but that's exactly when you need it most, because the debt accumulates fastest right there

precision is a discipline, not a default setting. every team I've worked with that ships reliable software has a process for it - code review, arch review, explicit constraints. 'let AI figure it out' skips all of that.

@rivestack docs are leverage most founders don't account for. "we'll write it later" - and then dev #3 joins and nobody wrote anything down

@a_protsyuk the "scales the team" framing is the real point. actual leverage is reducing how many things block on you. readable commits and clear docs buy back everyone else's time, that compounds

the 10x engineer myth is mostly wrong. what I actually see: engineers who make decisions once and document why. not the fastest coder. the one whose PRs read like a story and whose commits explain the context. that person scales the whole team.

@rivestack traction debt is the one that actually ends companies. tech debt slows you down. traction debt ends the story

@a_protsyuk traction debt is also way harder to see coming. tech debt shows up in your error logs. traction debt shows up in your churn rate six months later when you can't figure out why growth stalled

this. seen teams rebuild their entire architecture in 6 weeks. never seen anyone rebuild their distribution in 6 weeks. tech debt is recoverable. traction debt isn't.

@fadicheema12 @andrewchen right. you can automate the logistics. you can't automate "what are we actually hiring for and why". that conversation still needs humans in a room

@a_protsyuk @andrewchen Exactly! AI can optimize calendars and handle logistics, but if a team spends 60% of interview rounds just aligning internally, that’s a problem of clarity and decision-making, not automation.

everyone's talking about AI replacing middle managers but the real unlock for me would be for AI to: - do my email - replace the need for all my weekly meetings and 1:1s - make hiring loops more efficient 😂

@gauravtoshniwal half the product is underselling it. most AI integrations fail because context is wrong or missing, not because the model is bad

@a_protsyuk EXACTLY! Putting the knowledge base in .md files works for local and personal use cases. Capturing and passing relevant and limited context is half the product.

everyone's building multi-agent orchestration frameworks. we've been running 12 AI agents in production for 3 months. the boring truth: 90% of the work is context management, not agent coordination. your agents are only as good as the project knowledge they can access #AI

@hedgecast @housecor load-bearing dysfunction. the consultant who actually fixes it usually gets fired for it. seen it more than once

@a_protsyuk @housecor 6-12 months is optimistic. sometimes the dysfunction is the org's actual product and fixing it would break the politics holding it together.

The problem with an immature IT org is it tends to repel the people who are most capable of fixing it. When easy things are needlessly hard, and the org won't change, then the "A" players don't quietly accept it. They leave.

@joncphillips the review part is where judgment still lives. AI ships code fast - humans decide if it belongs there at all

@a_protsyuk Yeah for sure that part hasn’t changed, for me it’s really made everything faster. And I still review code and design systems.

Vibe coding has apparently ruined it for devs and software engineers. I’ve never written more code in my life, never shipped faster, never been busier.

@paulabartabajo_ lol exactly. AI-polished walls of text get ignored faster than unpolished ones. the problem was never the writing style - it's that nobody agreed on what needs to happen

@a_protsyuk They don't talk to each other, instead they dump their thoughts into Claude, copy past the beautiful report, send it on Slack, and expect the other side to actually read it xD Crazy times we live in

The CEO calls it AGI The ML engineer calls it RAG The data engineer calls it data plumbing

@rivestack exactly. and when that complexity tax hits, you're paying it while under feature pressure. worst possible time to refactor service boundaries

@a_protsyuk hard agree. split early and you're basically pre-optimizing for scale you don't have yet. the complexity tax shows up at the worst time

most startups split into microservices too early. monolith first. split when you know where the load actually is. i've cleaned up 3 of these this year - 4 engineers spending 60% on infra coordination, pre-PMF. the complexity costs more than it buys you.

@stanlee0nX syntax errors are the easy part - paste the error, fixed in 2 min. the 4 hours come from the model confidently building on a wrong assumption for 20 iterations while you don't have enough context to catch it

@a_protsyuk I don’t agree lol The problem is sometimes syntax, syntax joining one file to another

Average vibe coding experience: Spending 4 hours debugging with AI… a bug you could’ve fixed in 5 minutes if you actually knew how to code.

80% reduction in the grind is real. what's left is the 20% that actually matters - knowing which diagram is wrong, not just which one is missing. that judgment doesn't come from the tool.

the structure is clean but every one of those roles still needs someone making technical decisions. agents don't own trade-offs. who decided the data model? the infra approach? what gets cut? that's the layer that's missing from most one-person AI companies.

@FarzaTV second time is different in ways that matter. you know which problems are real and which are noise. you know how long the hard parts actually take. the scary part this time is you can't blame inexperience when things get hard.

I decided to join Y Combinator, again. This would be my second time! Not fully sure what I'm working on yet. But, I'm sure I'll find something in time as I wander and ship. I'm a little scared to do the whole build a company thing again ngl, but mostly excited. There's never been a better time to work on the ideas in my head. The batch started this week. Starting a company at 23 vs now starting a company at 30 feels so different. At 23 (when I did YC in 2020), naivety was there. At 30 I guess I know how difficult it all is. It's not surprising to me that most people in YC are aged 19-24. Still, I feel like I have the naivety of a 19 year-old, but, with the mental of a guy who's been through a lot and learned a lot. So, I'm bullish. Let's see what happens. You'll probably see me launching a lot of random stuff over the next few weeks especially. Also, I am blown away by the number of founders in the batch walking up to me telling me they credit being at YC to @_buildspace. It's so wonderful, and warms my heart. I often struggle to stop and understand the value of my past work because I'm so interested in the future. So, this was nice. It's funny, many saw me irl and freaked out thinking I was joining as a YC partner and were very very surprised to hear I was joining as a founder back in the dirt alongside them haha. Most founders never start another company and usually turn into VCs or get a high-tier job at a big company. I do not blame them. And honestly, that would be the easier more secure path for me especially as I begin thinking about family. But, idk. I feel like my ideas are important. And even though I don't have a specific "This is the idea I'm excited about" it's more a feeling of "I should explore my ideas...I would regret it if I didn't". Especially in 2026, at the epicenter of one of the greatest inventions of my lifetime. Every time I think about getting a job (of which I've been offered many great ones) that voice in my head comes back and says to give my nascent visions a shot. So, gonna try :) Maybe I flop, maybe I don't, only one way to find out. I'll be dropping weekly updates on YouTube if you're interested. I put one out last week that talks more in depth around the story of how this YC stuff even happened randomly, why I'm doing this again, my imposter syndrome and how I think about it, and other stuff. I'll link it below. Lets see what happens!! See y'all.

@jeffzwang google can't cannibalize their own products. gmail and docs generate too much revenue as-is. a truly useful AI assistant in them would reduce usage, not grow it. different incentive than anthropic or openai building from zero. structural problem, not a talent one.

It's honestly impressive how useless Gemini is in gmail, sheets, and docs, considering they are the perfect products for AI

@xoaanya bad execution on a good idea usually recovers. bad idea with great execution just gets you to the wrong place faster. but the real killer is neither - it's losing conviction before you get signal. most teams quit right before the thing would have worked.

What kills more startups: bad ideas or no execution??

@jonoringer the part nobody covers in these stories: someone made the core technical decisions early. which stack, which infra, what to build vs buy. that's not AI work. most 2-person shops i consult for stall not from lack of code but from missing that one decision-making role.

The NYT just profiled a $1.8B revenue company with 2 employees. Medvi is a telehealth GLP-1 provider built by Matthew Gallagher, 41, from his house in LA. He launched in September 2024 with $20,000. Here are the numbers: Month 1: 300 customers Month 2: 1,300 customers 2025 full year: $401M revenue, 250,000 customers 2026 run rate: $1.8B Net margin: 16.2% ($65M profit) Total employees: 2 (him and his brother) Outside funding: $0 How it works: Medvi is a front end. Two platforms — CareValidate and OpenLoop Health — handle doctors, prescriptions, pharmacies, shipping, and compliance. Gallagher handles brand, website, ads, checkout, and customer service. All built with AI. His stack: ChatGPT, Claude, and Grok for code. Midjourney and Runway for ad creative. ElevenLabs for voice. Custom AI agents to connect systems. AI chatbot for customer service (which initially hallucinated fake prices he had to honor). For comparison: Hims & Hers did $2.4B revenue last year with 2,442 employees and 5.5% net margins. Gallagher is running 3x the margin with a fraction of a percent of the headcount. Back into the unit economics: ~$336M in total costs, probably $160-200M to the telehealth platforms, leaving $130-170M mostly in marketing. Against 250,000 customers, that's a $500-700 CAC. High, but it works because his overhead is virtually zero and LTV at ~$200/month holds up. He's expanding fast. Men's health launched in February — 50K customers in month one. Meal delivery went live last month. Women's health, hair growth, supplements, and skincare are next. The vulnerability: zero moat. No proprietary tech, no doctor network, no pharmacy infrastructure. CareValidate or OpenLoop could raise fees or launch competing brands. Anyone could replicate this model in weeks. Right now, the margins are enormous for anyone who moves fast enough. The question is how long that window stays open. nytimes.com/2026/04/02/tec…

@0xrwu yeah. the tedium was hiding the real work - we were too busy writing boilerplate to think about design. now that's gone, what's left is the actually hard part: figuring out what to build and why. that part isn't getting automated.

Coding agents have really opened my eyes on how god dam tedious writing software was.

@jack used goose on a few client projects. the tool-use reliability is better than most. one thing i keep wanting: persistent state across sessions. right now you rebuild context every run - fine for one-offs, breaks for multi-step production workflows. hopefully team is on it.

people are sleeping on how excellent goose has become under the hood (interface needs some work but team is pushing). it's a superpower. github.com/block/goose