I went through ~3,000 lines of the new open-sourced 𝕏 feed code. Not out of curiosity. Out of wanting to stop lying to myself about how “smart” this thing is. Here’s what you actually want to know if the goal is growing efficiently without mystifying the system or coping about shadowbans. 👇

Hot take: The funniest thing about building a personal brand is that X is far from the best platform to achieve results People who grow fast use either vertical videos or streams So maybe we should just have more fun here instead of trying to be useful with every post? Day 4.

To be successful you need around 6-7 AI agents 😤

Every AI startiup that hit 1,000 users without ads did one of three things: 1. founder built an audience (and therefore, trust) before the product 2. the product created shareable output, so every result was a free ad 3. they manually onboarded the first 100 and asked each one to bring a friend Notice what's missing: launch week, press, paid KOLs Day 3.

profile visits is the main metric you need to care about not impressions

Your AI product doesn't need a marketing budget to get its first 100 users It needs 30 days of this: - find 20 accounts your users already follow - reply with actual insight, 15 times a day - post one specific lesson/vision piece daily, not product updates - DM everyone who engages with you twice and offer them a walkthrough - do this preferably from your personal account, not the project's one (people trust other people way more than logos) Boring. Manual. Works every time. Day 2.

I grew web3 projects from 10k to 100k followers. Now I'm doing it for my own account, from 1k followers, in public. Everything I learned running growth for 20 projects goes here. The playbooks, the numbers, the failures. Day 1.

The Hyperliquid trade isn't a crypto perp trade anymore. The 30-day data makes that hard to argue with: → HL processed $172.63B in 30-day perp volume, 31.9% of all tracked perp DEX activity → 24h open interest sits at $9.66B, BTC/USD alone doing $2.26B daily → HIP-3 OI hit a record $2.74B on May 27, up from ~$260M two months ago → RWA perps now account for 44% of total perp DEX volume on the platform A few things break when you run automation on HL. First, the market isn't crypto-only anymore. Oil, silver, and equity index perps now trade 24/7, with enough OI to support real grid and mean-reversion setups. During peak hours, WTI volume has flipped ETH on the platform. Second, funding moves faster than CEX-trained bots expect. HL settles every hour vs. 8h on most CEXs. For HIP-3 perps, a more responsive premium formula is used, capped at 4% per hour. That's actually less aggressive than many CEX counterparts despite sounding tighter. Setups copied from Binance behavior can bleed before the thesis plays out. Third, liquidity is concentrated where most retail bots aren't looking. Most automation we see is still focused on BTC, ETH, and HYPE, where the action used to be. If a bot has been running unchanged since Q1, it was designed for a different market. Range bounds, funding assumptions, and session timing all need another look. We've been rebuilding a few of our own grids from scratch on the new HIP-3 pairs. Mixed results so far. We'll share what's working in a follow-up.

#AI can be a valuable tool and, at the same time, it calls for a measured and vigilant approach. The speed and simplicity with which practical assistance can be accessed undoubtedly makes life easier. Yet they can also encourage excessive reliance and the search for ready-made answers, and weaken personal creativity and judgment. #MagnificaHumanitas

If we can suddenly field large numbers of high-level researcher minds, then the subfields that will be accelerated the most are those where cognitive labor can be applied at scale to already-available information. Where there is a fast and cheap verification loop, this acceleration compounds and things properly take off quickly. Where there isn't, the bottleneck simply gets moved. I was thinking about this after reading some of the responses to the GPT Erdos solution that talked about how attention-bottlenecked fields will see the biggest changes from attention-at-scale. GPT-5.5 Pro generated the image. Incredible improvement in text legibility, this would not have been possible a few months ago.

I lowkey might be the reason why people set restrictions on who can reply to their posts sometimes I just go too wild for the clients:)

making my acc less dead and more useful lately looking to connect with BD managers across web3 / AI / gambling / infra / exchanges / ecosystems have a nice partnership proposal drop a comment and let’s connect

feels good to be an alien these days

the hardest part of being a vibecoder so far is to spend a whole day without tweeting about AI even once

I'm starting from the basics littlemight.com/g-brain/

based side project for the weekend - status - found - ☑️

drake is a pure man in his 30s who's gotta flex his video production team for making so many videos with him proving he's young and still got it great job, creative marketing department

this

I am back.

attention to visuals is what make projects look outstanding in the current market great job!