Vladyslav Dalechyn

I had a dream today where I was using ChatGPT to generate videos of my two other dreams (in a dream) on-fly. AI bubble has another milestone to finish 😂

no fucking laptop costs $200B

$21.6B hyperliquid:native vs $1.75B solana:pumpCmXqMfrsAkQ5r49WcJnRayYRqmXz6ae8H7H9Dfn btw

MMM market making memes

have there been any improvements in code suggestion IDEs? i liked github copilot idea, but back then - it was unaware of the context (had single file context window if any). i believe developers, me including, are degrading due to our strive to become more efficient to be faster with the use of genai, but we all know the truth which is it is very frustrating and the dev velocity is too high to keep up with. i feel like there is some sort of middle ground that is hugely unexplored and it is a problem of millions of developers. like instead of trying to go 100x, let's try going 3x. for quick projects surely it's like a drag strip - let's go as fast as we can. but for big projects - we need to tune ourselves but also match our expectations. try taking a drag car to a track - that's what happens with us most of the time. we need AI to boost us without the cost of code context.

we got larp VCs before GTA VI

top niches to be in future: - having an opinion - thinking and reasoning execution will be solved. thinking won’t be.

i am early testing boski, not very actively chatting with him, i'm not a huge fan of "chatting" overall, but boski is the first agent that really felt personal. when he pings me, it feels like a human. somewhat as if I missed my friend's message. it's worth a try, believe me.

a month ago, i resigned from the startup i spent most of my adult life building. i retained my shares, but i’m no longer involved in the company or its operations. Kien and i started building together when i was 19 after meeting at a Ukrainian hackathon. our first public product was @beliefssocial (beliefs.social), a memecoin launchpad on Base. the product was gaining traction, but slowly. then, in a single day, we built a Twitter bot that allowed people to launch tokens directly from their feeds. it brought a huge wave of attention and activity to the platform. after around two weeks, the hype faded, and we returned to roughly our previous level of traction. that experience taught us how closely launchpads are tied to the narratives driving the market. they can grow extremely quickly while a narrative is active, but when attention shifts elsewhere, sustaining that growth becomes much harder. we wanted to build something less dependent on short-lived narratives and more capable of generating consistent revenue, so we moved on to @tradeonmidas midas.trade, a memecoin trading platform that aggregates tokens from different launchpads on Base. we chose the technically hard route. we indexed the blockchain ourselves, ran our own infrastructure, processed terabytes of data, and built custom analytics on top of it instead of being limited by third-party providers. i could call myself the CTO, but technical co-founder is more accurate. the biggest engineering team i managed was four people. most of the time, it was just me and one other engineer. Andrii and i built most of midas.trade with our four hands. i’m proud of the indexer, the infrastructure, the UI we built without a proper designer, and the fact that a tiny team managed to build something this complex. but the honest reason i left is simple: i was tired. i spent years assuming i could push myself through anything. eventually, i realized that i had stopped listening to myself and that i was deeply burned out. the biggest lesson wasn’t technical. you can rewrite a system, fix infrastructure, or ship another version. alignment, communication, and working relationships are different. they need to be taken seriously from the beginning. i’m not angry, and i don’t regret it. every day i stayed was also a choice i made. i’m responsible for the choices that hurt me, but i’m also grateful for everything they taught me. @FamKien , Andrii, and @asherkrooz all fought for the same goal in their own ways, and they deserve to be thanked for it. right now, i’m recovering and doing a short DeFi engineering project around solvers and delta-neutral inventory management. i don’t know what comes after that. maybe more DeFi. maybe consumer crypto. maybe something completely different. for the first time in years, i’m comfortable not knowing. if you’re building something interesting, went through something similar, or simply want to talk, my DMs are open.

codex cli is so shitty compared to claude. the scrollbar is lagging, and i can't even see the model's response. if you switch to gpt, bring up opencode, although i'm not a fan of that UI either. is there claude-like CLI for gpt?

i replaced boring claude tips with hackernews articles and it's been the most interesting small project i've done for a while