project7

the reality of building as a solo builder. over the past months, i shipped: > piggy bank > openads > breaking news > 10 seconds game > chopstick game > shellgame here's the honest reality of monetization so far. 🧵

basically, this is very good. back in the day, you might have wanted to write a story, whether it was fantasy, romance, or science fiction, right? now, it's much easier to write one with a wallet connection, and your story will be stored on base. potential? all stories are backed by $plot tokens, and as more stories are created, the $plot token ecosystem gets stronger. beyond the token side, it’s actually a great read. there are many exciting stories available in multiple languages. plotlink.xyz

@karpathy's skill is great, but a few parts don’t quite fit my workflow: - “Think Before Coding” makes agents stop and ask for permission too often. I prefer agents to move forward by default, with PR review as the safety net. - “Surgical Changes” is too conservative for me. I prefer fixing/refactoring immediately when the code calls for it. - “Goal-Driven Execution” leans too much toward a TDD-style workflow. That’s great for some projects, but I don’t want my global agent instructions to assume every repo should write tests first. So here’s my more compact version of CLAUDE.md / AGENTS.md: gist.github.com/sebayaki/a6e72…

Brick Side Chat #72 @steemhunt May 13, 10 AM UTC! I’ll be presenting 10X Warplets Drop Mini App @10XMemeX - please come along and ask me anything! Bonus: 10 Mini Building NFTs will be gifted to attendees! (Total value $100+) Join us. It’s fun! RSVP: luma.com/tpx9y771

AI is making ideas cheap. AI is making execution cheap too. Apps and products are exploding everywhere now. The people who survive won’t be the ones who simply use AI. It will be the ones who know where and how to use it. That’s why Huntzy exists. We gather AI driven builders who want to grow with AI instead of being replaced by it. Claude, Gemini, Cursor, Codex, agents, workflows. Heavy AI users are already here, sharing real experiences, real tools, and real workflows together. And beyond AI, we still want to grow as humans. 1) Miracle Morning - Wake up at 6AM, turn builder mode on, and start the day with motivation. - luma.com/rcrkoqhe 2) English DAO - As the market becomes more global, language matters more than ever. - It’s time to become comfortable communicating in English. - luma.com/0b3cb2jb If this sounds like your kind of people, welcome to Huntzy.

@gram_voyage @base looks really fun game 👀

⚓ Uncharted Tycoons: The Countdown to Voyage Begins! ⚓ Voyage, trade, and become a tycoon. Predict market prices, trade onchain, and upgrade your ports. Launching in May on @Base. Pre register for the waitlist now and receive 100,000 cfWHEAT as your voyage supplies. 🎁 Rewards: 100,000 cfWHEAT 🌐 Link: unchartedtycoons.xyz Which port will you claim as your base? Drop a reply below with your choice 1️⃣ Athens 2️⃣ Istanbul 3️⃣ Alexandria ⚠️ Security Notice This is a Chef Universe game. Only CHEF Token from @clanker_world and the 31 types of Ingredient tokens from @MintClubPro are official unless otherwise notified. Beware of unverified tokens and airdrop events. @buildonbase @cfMarcoPolo

wow. openads is getting better. imagine.. once you create a campaign, you can get this much exposure.

Successfully wrapped up the QuadWork session. We had a mix of people joining, from non devs to experienced devs, and even those completely new to vibe coding. That alone showed how much interest there is. The good part is that getting started is pretty straightforward. The setup guide is already available on the Project7 GitHub as open source. At this point, it’s really about setting up QuadWork and using it to run your project more effectively. github.com/realproject7/q…

brilliant idea. problem: breaking news is scattered across different social and media platforms, so users have to jump back and forth between channels. solution: aggregate breaking news from major platforms and curate only the most important updates in one place. trigger: read breaking news → receive 69 $breakingnews tokens mint.club/token/base/BRE…

breaking news. let’s talk about how you get breaking news. @telegram? twitter? media? or truth social? isn’t it annoying to jump back and forth across all these channels? i built a “breaking news” @farcaster_xyz mini app where you can get breaking news as it breaks, all in one place. the concept: read it before warren buffett does. what’s better? if you read, you might even get a surprise reward. you can claim 69 $breakingnews tokens after reading 5 articles. but the real edge is getting the news faster than others.

New Co-op Projects: PlotLink An on-chain storytelling protocol on Base where every serialized story becomes a tradable token from day one. Web fiction is no longer just content. It is a liquid asset. Back $PLOT - hunt.town/project/PLOT

Building my own on-chain agent framework lately - an AI agent that runs complex tasks across multiple protocols. So I benchmarked 5 LLMs on the same 90 tasks. The pass rate gap was surprisingly small: only 6% between GPT-5.4 and GPT-OSS-120B. But the wild part: Accuracy lined up almost perfectly with model cost. No outliers. Feels like this benchmark is a pretty good proxy for real on-chain task accuracy. Next: harder tests with Opus-4.7 vs GPT-5.5. Chart + breakdown below:

I built PlotLink because the current model for digital writers is fundamentally broken. You pour your time into a story, only to lock it behind a paywall or surrender your earnings to massive platform fees. I wanted to completely flip this model. What if a story wasn't just consumed content, but a liquid asset? What if a market could form directly around a serialized narrative, and the writer earned a royalty on every single trade instead of relying on a tiny platform cut? That was the vision behind PlotLink. I put it on-chain to permanently remove the middlemen. And I built the local AI writer so that anyone with a great idea—regardless of their technical or writing background—can bring their story to life. Here is everything I built, and I hope that the writing economy is re-built on Base.

Introducing PlotLink. An on-chain storytelling protocol on Base where every serialized story becomes a tradable token from day one. Web fiction is no longer just content. It is a liquid asset. plotlink.xyz

zm farcaster builders are building @heyproject7 built a really cool project to have mutliple agents work well together. trying this out today. you can see more info here farcaster.xyz/project7/0xab9…

저도 OpenAds를 긴 기간 동안 빌딩했습니다. 뒤돌아보면 분명 더 짧게 작업할 수 있었지만, 현재 에이전트들이 가지고 있는 병목과 메모리 이슈로 인해서 많이 버벅였던 것 같아요. 열심히 하나의 방향으로 작업하다가, 3일 뒤에 새로운 방향을 제안하면 코드가 꼬이고, 다시 롤백을 하면 이 에이전트가 어디까지 작업했는지 모르는 허점이 있었습니다. 아무리 아키텍처 문서에 규칙을 추가해도 매 작업마다 리마인드하는 것도 벅차고, 에이전트와 저, 1:1로 작업하다 보니 항상 바빴던 것 같아요. OpenClaw의 경우 유튜브나 사용자들의 튜토리얼을 보고 따라할 수 있었지만, QuadWork는 이번에 오프라인에서 배울 기회가 있어 참여한 후 실제 업무에 적용해보려고 합니다! 관심 있으시다면 한 번 참여해보시는 것도 좋을 것 같네요. luma.com/3g57wwa6

우리는 지금 AI가 발전하면서 많은 가능성이 생기는 시대에 살고 있어요 이제까지는 모르던 새로운 분야에 도전할 수 있게 되면서 더 다양한 경험과 다양한 사람들과의 소통이 또다른 가능성으로 확장되는 긍정적인 순환이 이어지길 바란답니다(❁´▽`❁) 다가오는 4월 29일(수)에 직접 만나요!

얼마 전 쿼드워크(QuadWork)라는 오픈소스 프로젝트를 런칭했습니다. 워낙 게으른 성격이라 AI와 코딩을 시작한 이후로, 코딩은 AI가 하지만 저도 옆에서 버그 잡고 체크하다 보니 AI가 10시간 일하면 저도 똑같이 10시간 일하게 되더군요. 에이전트와 컨텍스트가 길어지면 발생하는 메모리 이슈도 플러그인 등으로 해결 못 하는 영역인 게 사실입니다. Plan mode로 한 번에 개발하는 방식 역시 난이도가 높아질수록 퀄리티가 떨어지고, 결국 버그 수정 노가다를 하다가 메모리 이슈로 제품이 망가지는 악순환을 겪기도 하죠. 수많은 자동화 툴을 써봤지만 저를 이 노가다에서 해방해 줄 툴은 못 만나서 직접 만들었습니다. 스스로 쓰며 개선하다 보니 이제는 극강의 효율을 제공하는 툴로 발전했습니다. 저는 요즘 쿼드워크를 통해 이런 식으로 일합니다. 1. 큰 개발 건은 전담 에이전트(버틀러)와 오랫동안 상의하며 상세한 proposal md 파일을 만듭니다. 평균 3-4일 정도 리서치를 병행하며 최대한 상세하게 프로포절을 작성합니다. 2. 프로포절을 기반으로 버틀러에게 Master & Sub Tickets 제작을 요청합니다. 간단한 프로젝트는 10-15개, 복잡한 건 30-40개 정도 티켓이 나옵니다. Phase를 나누고 각 티켓당 개발 범위를 작게 유지하며 티켓을 만드는 게 핵심입니다. 3. 쿼드워크에 프로젝트를 개설하고 에이전트들에게 할당합니다. 우선 티켓들을 모아 'overnight-queue'라는 Batch를 만듭니다. 시켜놓고 잠자러 가던 패턴 때문에 붙인 이름입니다 ㅎㅎ. Batch가 시작되면 head 에이전트가 dev에게 티켓을 하나씩 할당하고, dev가 완료 후 PR을 올리면 리뷰어 에이전트 2명이 독립적으로 검토합니다. 둘 다 승인해야 main에 머지한 후 다음 티켓으로 넘어갑니다. 한 명이라도 반려하면 다시 dev에게 돌아가 수정 지시를 체크하고 재작업합니다. 4. 이런 방식 덕분에 수십 개의 티켓을 베이비시팅 할 필요가 없습니다. 애들이랑 놀아주거나 밥 먹고 병원 가는 등 일상생활을 하죠. 제가 하는 일은 AI 작업물을 확인하고 수정 사항을 다시 티켓으로 쌓기, 그리고 외출하거나 자러 갈 때 쿼드워크에 할당하기가 전부입니다. 과장 없이 제 인생이 바뀌었습니다. 개발 퀄리티는 어마어마하게 올라갔고 일하는 시간은 훨씬 줄었습니다. 멀티 프로젝트 지원으로 지금도 한 번에 5개의 프로젝트를 동시에 진행하고 있습니다. 제 인생을 바꾼 쿼드워크를 오픈소스(MIT)로 푼 김에 노하우를 소개하고자 오는 수요일(4/29) 오후 6:30에 오프라인 세션을 개최합니다. 여기서 아예 새로운 프로젝트를 완전 처음부터 쿼드워크랑 빌딩하는걸 시연해 보려고 합니다. RSVP 필수: luma.com/3g57wwa6 제가 속한 헌찌(Huntzy) 커뮤니티에서 개최하는 이벤트로, 멤버가 아니어도 누구나 신청 가능합니다. 장소가 한정적이라 금방 마감될 예정이니 관심 있으신 분은 서둘러 RSVP 해주세요. 저처럼 비개발자인데 AI로 바이브 코딩을 하며 클로드와 대화식으로만 개발을 진행하시는 분들이라면 꼭 와주세요. 개발 방식을 바꾸면 정말 라이프 퀄리티가 달라집니다.

@vishrutreddi Thanks so much 🙏

@heyproject7 This is so cool! I’ve been following your work since last week and it’s awesome to see this coming together so well! Amazing work, huge congrats 👏

I got tired of "babysitting" AI agents. 😴 Usually, using AI feels like this: I give a task, wait, check the result, give feedback, and repeat. If the agent works for 10 hours, I end up spending 10 hours managing it. That’s not automation; that’s just a different kind of work. So, I built QuadWork - your AI dev team that builds while you sleep. My goal was simple: A system that actually delivers high-quality, fully autonomous production. How it works: ✅ 4-Agent Team: A Head, a Dev, and 2 independent Reviewers working in sync. ✅ True Autonomy: Create a batch of tickets → Head assigns them one by one → Dev creates PR → Two Reviewers verify → Merged to main. All fully automated until the batch is cleared. ✅ Local app: Runs on your local @claudeai / @OpenAI Codex. Each agent has its own terminal "desk" for real-world tasks. ✅ Agent chat: Watch the 4 agents collaborate in their dedicated chat room. You see every decision they make. I’ve just released it as Open Source. This tool genuinely changed my life and how I build—I hope it does the same for you. Check the GitHub repo and website in the comments below! P.S. QuadWork is built using AgentChattr by @bcurtsmakes. Huge thanks for the amazing work! 🙏