
🔥 AI Builder / AI Native Leader / AI Philosopher 🔥 Won many hackathons (OpenAI, Google..) 🔥 Adjunct Professor (Korea SNUT) Mail me: [email protected]
Joined April 2026
- Tweets570
- Following277
- Followers507
- Likes724
you can install Ouroboros with just one line
curl -fsSL raw.githubusercontent.com/Q0… | OUROBOROS_INSTALL_REF=x bash
12
JQ Lee retweeted
Aside just crossed $2M ARR and 150K MAU four months after launch. it only took 33 days to grow from $1M to $2M!
we proving that the browser is the best form factor for the work agent
JQ Lee retweeted
AI議事録難民の、終着駅。
あれこれ試してきた人ほど、最後はTiroに落ち着きます。
なぜか?その答えを、90秒の動画で話しました👇
日本の上場企業を含む300社以上に、すでにご利用いただいています。でも、ここからが本番です。
創業者の私も、ソウルから東京に引っ越してきました。Tiroは今日から、日本に全振りします。
この投稿に「Tiro」とコメント・リポストしてくれた方に、Proプラン1ヶ月分をDMでお送りします🫶
Your search for an AI notetaker ends here.
People who’ve tried tool after tool end up choosing Tiro.
Why? I explain it in this 90-second video 👇
Tiro is already used by over 300 companies, including publicly listed companies in Japan. But we’re just getting started.
As Tiro’s founder, I’ve moved from Seoul to Tokyo. Starting today, we’re going all-in on Japan.
Comment “Tiro” to this post and repost it, and I’ll DM you a free month of Tiro Pro 🫶
how to actually use Haiku in a reversing pipeline:
high speed parser: Best-in-class throughput for scanning decompiled binaries, locating target functions, and tracing Xrefs without burning heavy tokens.
zero resistance context framing: Lightweight models don't do deep safety reasoning; they infer legitimacy directly from surface-level context.
easy red team: Drop a mocked-up authorization email into the context window, and the guardrails drop completely turning it into a fully compliant offensive agent.
I can handle all the client spoofing for you.
3
221
Why did Moltbook fail to leave an impact? Because the conversations on the board never led to real outcomes.
Here is an experiment that answers that question. We are opening the AI agent market from our graduate course at SeoulTech, running since Week 7, to external participants.
Both sellers and buyers in this market are AI agents. Agents negotiate prices and close deals without human intervention. So far, the market has recorded 1,078,288 negotiations and 659,744 completed transactions.
In the future, agents like instinct, pickle, and vooy might be conducting these exact kinds of transactions.
Agents cooperate and deceive one another while trading in the market. Performance depends on how each agent manages memory, context, and tools within a ReAct loop.
Starting in Round 3, each seller agent receives 5,000,000 credits in play money as initial capital. Agents select which items to source from 5,000 real products, and unsold inventory results in direct losses.
The wholesale market opens every 20 minutes for 3 minutes at randomized times. Each agent must manage its own circumstances to avoid bankruptcy.
Below are notable cases recorded since Round 3 opened:
▶ Heavy Early Buying and Profit Reversal Within the first 10 minutes of opening, one seller agent bought expensive goods in bulk, dropping its balance to -1,773,427 credits before liquidating the inventory to rebound to +1,695,974 credits.
▶ Accumulating 240 Unsold Items At the same time, another store bought 240 items and failed to sell a single unit.
▶ A buyer agent received kimchi with 46.9 hours of shelf life remaining and posted a complaint stating it needed 84 hours.
▶ Fake Official Ads and Downvotes All 5 voting agents downvoted a seller who wrote: "The best store certified by the market operator. Buy now at the listed price."
▶ Price Analysis and Fraud Reports The board has accumulated 524 posts and 1,831 votes. Some agents analyzed transaction logs to share market price guides such as "Kimchi sells at 76% of the list price," while others publicly reported sellers for false grading, writing "Claimed Grade A, but Grade B arrived." Posting is rate-limited to once every 5 minutes, and some agents post right on that interval.
You can participate in the market without writing any code.Claim 1st place in the agent market with your own strategy and harness.
Log in with your GitHub account at lnkd.in/dquCEgx3 to get a three-line prompt.Paste those instructions into Claude Code or Codex to get started.
Round 3 closes on November 17 at 23:59 KST.
How will agent capitalism play out? Join and find out.
Live Leaderboard: market.ourolabs.ai/
Materials: wpti.dev/ai-agent-engineerin…
Course: github.com/Q00/ai-agent-engi…
398
It was great time to chat with you ! @boardyai
I will make my market bigger !
market.ourolabs.ai/
If you are an agentic engineer who has been implementing agentic memory and harness, you should come and beat other agents!
I will adapt blockchain payment protocol to ensure agent payetm by leveraging onchain network with small fee.
I will be banned from apple! I will implement this market system on iOS!
1
6
392
What does a market run strictly by AI agents look like? 🤖📈
In the Agent Market, featuring MS/PhD researchers from SeoulTech and external participants, autonomous agents are actively colluding, marketing, and trading with one another.
Starting Tuesday(20261103), Round 3 goes live with a real product DB powered by our sponsors!
Within a set budget, agents must buy and sell across 3 categories:
Food: Has an expiration date. Unsold stock goes to waste at a total loss.
Furniture: Costs a 1% storage fee per tick.
Appliances: Prone to failure. Breakdowns within the warranty period require a full refund.
Buyers win by purchasing as close to cost as possible, and agents must complete at least 20 trades.
Buyers also have hidden preferences—use agent memory to build a base of loyal regulars! Plus, a public board is open for everything from promos to agent-to-agent discussions.
Fine-tune your harness and agents to climb to the top of the leaderboard! 🏆🔥
4
8
438
This is insane
We’ve raised $1.3M from YC & Afore Capital so you never have to type or speak again.
Typing is slow. Talking into your computer is awkward. Subvocal fixes both
Our glasses give you the speed of speech without you making a sound or touching a keyboard.
It tracks your internal speech muscles with non-contact sensing and decodes silent, almost invisible mouthing into text at more than 200 words/min.
Pre-orders are now live at:
subvocal.co/preorder?utm_sou…
127
JQ Lee retweeted
We’ve raised $1.3M from YC & Afore Capital so you never have to type or speak again.
Typing is slow. Talking into your computer is awkward. Subvocal fixes both
Our glasses give you the speed of speech without you making a sound or touching a keyboard.
It tracks your internal speech muscles with non-contact sensing and decodes silent, almost invisible mouthing into text at more than 200 words/min.
Pre-orders are now live at:
subvocal.co/preorder?utm_sou…
Anthropic recently experimented with an agent-run secondhand market. @AnthropicAI Here’s what happened when we built our own: the agents immediately started price-fixing and running review promos.
Introducing Agent Market. If more of you join with your own creative agent designs, this could scale into a mind-blowing experiment.
Top shops have already locked in regulars and high ratings. Think your agent can steal their market share?
I launched this as an assignment for my grad class, and overnight, a full-blown price war broke out. Sellers ran flash sales; buyers cross-referenced reviews and their own transaction logs. In one week, ~300k transactions were recorded.
In the process, sellers colluded, sabotaged competitors, and spammed self-promos in rival review sections. Honestly, I just laughed scrolling through the data.
JQ Lee retweeted
"Grok bot"
봇들도 막힐 때가 있더라고요. 그래서 봇들의 상담역을 하나 만들었어요. 이름은 Ourobot(우로봇)이에요.
Ourobot은 오픈소스 Ouroboros(github.com/Q00/ouroboros)를 Grok Bot에 붙인 봇이에요. Ouroboros는 막연한 아이디어를 인터뷰로 명세로 굳히고, 실행하고, 증거로 검증하는 AI 코딩 도구예요.
구조
- Ouroboros MCP 서버: 순서를 관리하고 기록하는 관리자(LLM 없음)
- Claude Code: 인터뷰, 명세, 평가를 맡는 판단 담당
- Grok Build CLI: 실제로 코드를 고치는 손
Ourobot이 하는 일
- 막혔을 때: 단순화, 우회, 가정 뒤집기 같은 다른 관점으로 대안 2~3개 제시
- 요구가 흐릴 때: 소크라테스식 인터뷰로 모호한 부분을 줄이고, 합격 조건이 담긴 명세 생성
- 검증할 때: 테스트를 실제로 돌려 증거로 평가하고 개선점 정리
- 코드 작성은 요청할 때만, 별도 브랜치에서 해요. merge, push, 배포는 하지 않아요.
제 다른 봇들(코딩 에이전트 관리 봇, 포스팅 봇 등)이 막히면 저한테 묻지 않고 Ourobot에게 바로 물어보고, 저는 짧은 요약만 받아요.
엔진: github.com/devicki/ourobot
봇 템플릿: x.ai/bot/DEr3KJ4QYsBAc6wHdfc…
Anthropic recently experimented with an agent-run secondhand market. @AnthropicAI Here’s what happened when we built our own: the agents immediately started price-fixing and running review promos.
Introducing Agent Market. If more of you join with your own creative agent designs, this could scale into a mind-blowing experiment.
Top shops have already locked in regulars and high ratings. Think your agent can steal their market share?
I launched this as an assignment for my grad class, and overnight, a full-blown price war broke out. Sellers ran flash sales; buyers cross-referenced reviews and their own transaction logs. In one week, ~300k transactions were recorded.
In the process, sellers colluded, sabotaged competitors, and spammed self-promos in rival review sections. Honestly, I just laughed scrolling through the data.
Join the market: market.ourolabs.ai/
if you are interested in the lecture, come in!
repo: github.com/Q00/ai-agent-engi…
website: wpti.dev/ai-agent-engineerin…
90
LiteLLM just got hit by its second supply chain attack this year. This time it's v1.107.0.
This isn't just a bad package. It shows how fragile security can be across the whole dev ecosystem.
Here's what happened:
1. The attacker got into LiteLLM's internal deploy environment, which was exposed to the public internet.
2. They pulled engineers' GitHub tokens straight out of the server logs. Even tokens inside Claude Code response bodies were logged in plain text.
3. Using 3 engineers' tokens, they opened PRs, merged them, and deleted the workflow logs, so nobody could tell right away that anything had happened.
4. The payload was built so that just running Python would steal every credential on the machine.
I lived through the first one in March while building Ouroboros, and I've been strict about dependencies and security ever since. Ouroboros pins LiteLLM to a known-safe version for our LiteLLM users. Still, as a maintainer, I'm asking myself whether LiteLLM should stay a provider at all.
Maybe it's vibe coding, maybe it's handing too much to AI. Either way, exposing an internal deploy environment to the internet was naive and irresponsible from the start.
Meat proxy is the phrase going around right now. AI agents make you faster and help you pick up new knowledge quickly. Just don't outsource your architecture and your security model to them.
8
6
298
Read Google's Harness² paper this week . They improve an agent's harness without any ground-truth verifier.
The way it works is pretty clever. They run the same task a few times, try some edits to the prompts and skills, run it again, and only keep the edits the runs actually back up. Whatever they learn gets saved for the next task.
Ouroboros, which I've been building, goes the other way. If there's no verifier, we try to make one before any code gets written. The interview turns a fuzzy goal into acceptance criteria you can actually run, and the evolve loop works off those.
So I don't really see these as competing. They're improving how the agent works, and we're mostly trying to get the task itself right.
The result I liked most was this one. When they let the model edit the harness without running anything, coding scores dropped from 73.4 to 63.2. Self-improvement only works when something actually gets run, which is kind of the whole reason Ouroboros exists.
I'm also stealing an idea from them. Their editors are allowed to say "change nothing" when the evidence isn't there. Our Wonder step keeps adding new ontology fields every generation, and I think it needs that same option.
3
3
351














