
@tradecatlabs 研究员 构建:tradecat、fatecat、Human Infra、vibe-coding-cn、vibe-mathing-cn、prompt_jsonl 目标是实现有效永生和记忆编辑 🤖
Joined June 2022
- Tweets8.2K
- Following4K
- Followers14.1K
- Likes30.1K
123olp retweeted
币安被盗那一个小时,链上出现了一个教科书级的套利窗口。有人一小时赚了 347 万刀,我把链上数据扒了一遍,复盘一下。
【发生了什么】 北京时间 10 月 30 日凌晨 4 点 24 分左右,币安热钱包被盗。 黑客手里一大堆 USDT,怕被 Tether 冻结,就在链上 DEX 疯狂把 U 换成 ETH。同一时间币安暂停提币,交易所里便宜的 ETH 搬不上链。 结果:
Uniswap 上 ETH 在 04:40–05:40 被买到 2560–2750,个别成交 3000+
Aave 用的 Chainlink 预言机汇总的是交易所价格,全程在 2250–2290 链上比交易所贵了 15%–30%,挂了将近一个小时。
【套利连招】
往 Aave 存 USDC/USDT 当抵押
按预言机价 ≈2280 借出 ETH
马上去 DEX/聚合器按 2600+ 卖成 U
卖出来的 U 再存回去,继续借、继续卖
价格回落到 2290 附近,买回 ETH 还款,差价全是利润 这不是黑 Aave,Aave 合约一点问题都没有。本质是 CEX 和 DEX 的价差搬砖,Aave 只是一个"按交易所价格出 ETH 的窗口"。
【链上数据】我拉了事发后 6 小时所有在 Aave 主网借 ETH 的地址,一共 61 个,排除黑客地址(6000 万 U 抵押借了 19762 ETH),逐笔算了盈亏:
第一名 0x0cdd…61D8:卖出 12289 ETH,均价 2580;买回 12314 ETH,均价 2297;▶赚约 347 万刀◀
第二名约 78 万,有一批卖在 3000 以上
第三到五名分别约 67 万、43 万、29 万
前十名合计约 636 万刀 206 笔卖出平均比预言机高 10%,最高超过 30%。
第一名的操作很值得学:
事发后 24 分钟就进场
Aave 里早就存好了 3700 万 USDC
分 18 笔卖,每笔 300–1000 ETH,用聚合器拆单降滑点
价格一回落就开始买回还款,两个小时内基本平仓
【学到的】
CEX 出事、暂停提币的时候,链上和交易所的价差会挂很久。平时几秒钟就被搬平的价差,这次挂了一个小时。
机会在那些按交易所价格定价的地方:用 Chainlink 预言机的借贷池(Aave、Compound、Spark……),它不跟 DEX 现价。
弹药要提前放好。等你看到新闻再去充 U、开仓,窗口早没了。
窗口很短,DEX 越卖越便宜,能赚多少看深度和速度。05:30 以后价差只剩 3%–5%。
Aave 很快冻结了主网 V3 市场,晚进场的连借都借不到。
【风险】
借 ETH 的仓位怕 ETH 继续往上冲,健康度会掉,所以大户都分批做,不会一把梭
没买回的 ETH 负债还挂着,价格回升就会回吐利润
链上拥堵时 gas 很贵,小资金可能做不划算
下一次黑天鹅来的时候,先想清楚自己的"出 ETH 窗口"在哪里。
(数据:链上逐笔统计,按 ETH 2280 计价,未扣 gas。不构成投资建议。)
人类所设定和追求的“完美状态”,本质上是通过想象与叙事建构的理想图景,并无与之对应的既定客观现实。正是这一虚构图景在观念中的形成,使尚未实现的状态得以被确立为可追求的目标,进而形成对未来的预期、引导现实行动,并为目标的实现创造可能。这与尤瓦尔·赫拉利关于人类“讲故事”能力的论述相呼应:人类能够构想现实中尚未存在的事物,通过赋予这些想象以意义,使其成为认知、预期与行动的依据,并以此推动想象中的可能状态向现实转化。
#vibecoding方法论,(root)固定点收敛系统可界定为自主智能体的分层闭环控制框架,用于将意图转化为可执行、可验证且受变更约束的目标状态,并持续缩小现实与目标的偏差。其机制遵循“观察→差距识别→行动→验证→修正”,行动失效则升级策略,策略失效则复审目标,目标不可确定时回退需求或交由人工裁决。该框架依赖目标可识别、状态可观测、验证准则一致及策略空间中存在有效控制路径;缺乏稳定性或充分反馈时,“收敛”只能作为架构期望。方法上形成“需求澄清→目标冻结→计划→执行→验证→更新→升级”的闭环,并与需求工程、反馈控制、自治计算、监督控制及运行时保障相衔接。其核心问题是:动态环境下,如何以分层约束获得对目标趋近过程的可验证控制。
123olp retweeted
自己跑Ubuntu的,9.8高危漏洞,速度打补丁。
Linux 内核高危漏洞 CVE-2026-98432。
问题出在 IPv4 TCP 网络栈:并发处理连接请求时,内存可能在仍被使用的情况下提前释放,形成 use-after-free。
Anthropic 声称其未发布的 AI 模型将该漏洞串联成远程 root 命令执行攻击,约 53 分钟完成利用验证。
# 查看系统版本
cat /etc/os-release
# 查看当前内核
uname -r
升级 补丁
sudo apt update
sudo apt full-upgrade
sudo reboot
参考:cve.org/CVERecord/?id=CVE-20…
123olp retweeted
this might be my last normal tweet before biotech goes vertical
i'm not a whistleblower. I don't work at Lilly, Novo or any of the big labs
but there is a massive shift happening in human biology right now.
it's sitting in press releases and almost nobody reads them
the last 6 months alone:
- a once-daily pill roughly doubled median survival in metastatic pancreatic cancer (13.2 vs 6.6 months)
- a one-time gene editing infusion cut hereditary angioedema attacks 87%, and 62% of patients stayed attack-free and off all other prevention for 6 months
- a one-time gene edit for cholesterol dropped LDL up to 62%, and it held for over a year
- a personalized mrna vaccine beat keytruda alone in a 1,137 patient melanoma phase 3
- retatrutide hit ~28% weight loss in a phase 3
- an ai-designed drug entered phase 3
- a pill that regrows hair passed its second phase 3
- an immunotherapy that could let some rectal cancer patients skip surgery, chemo and radiation just got fda approved
- a man lived 271 days off dialysis on a pig kidney
- human cells were caught passing chromosome-sized chunks of DNA to their neighbors, and it changes how the neighbors behave
we are on the verge of an era where the human body is a software problem and the update cycle gets shorter everyday
stick around, it's going to get weird
bio/acc
If you look closely enough, breakthroughs in everything are accelerating exponentially.
An exponential always starts off looking small.
this might be my last normal tweet before biotech goes vertical
i'm not a whistleblower. I don't work at Lilly, Novo or any of the big labs
but there is a massive shift happening in human biology right now.
it's sitting in press releases and almost nobody reads them
the last 6 months alone:
- a once-daily pill roughly doubled median survival in metastatic pancreatic cancer (13.2 vs 6.6 months)
- a one-time gene editing infusion cut hereditary angioedema attacks 87%, and 62% of patients stayed attack-free and off all other prevention for 6 months
- a one-time gene edit for cholesterol dropped LDL up to 62%, and it held for over a year
- a personalized mrna vaccine beat keytruda alone in a 1,137 patient melanoma phase 3
- retatrutide hit ~28% weight loss in a phase 3
- an ai-designed drug entered phase 3
- a pill that regrows hair passed its second phase 3
- an immunotherapy that could let some rectal cancer patients skip surgery, chemo and radiation just got fda approved
- a man lived 271 days off dialysis on a pig kidney
- human cells were caught passing chromosome-sized chunks of DNA to their neighbors, and it changes how the neighbors behave
we are on the verge of an era where the human body is a software problem and the update cycle gets shorter everyday
stick around, it's going to get weird
bio/acc
123olp retweeted
“预计AGI奇点在xx年后到来”,这种说法是不准确的,不同职业不同阶层的人进入AGI奇点的时间不同
对于数学、理论计算机科学的许多科研人员来说,2026年10月他们已经进入AGI奇点,他们是最早跌落AGI奇点的人类
未来如果有历史学家试图界定人类社会最早接触并跌落AGI奇点的时间,我提名,2026年10月
如果你想知道进入奇点的那种眩晕、极度不真实、诡异到恐怖的末日感
可以去问问Terence Tao、王虹等人,因为他们的主要专业领域——经典调和分析,最顶级的四大猜想已经被OpenAI一举荡平了,但新的人类理解还没形成,新的深刻问题还没出现,简直就是被OpenAI瞬间核平后的废土世界
也可以去问多数工作基于ETH/SETH等这个月被推翻的假设的理论计算机科学家,他们很多人一辈子的工作刚被OpenAI宣布是彻头彻尾的幻觉,很多博士生手头关乎毕业生计的论文还没拿到credit就被宣布无效
不同职业、不同阶层的人,离AGI奇点的距离,掉落AGI奇点的速度是不同的,纯度越高的脑力劳动,世俗意义上越是“聪明人”、“学霸”从事的职业,越是先跌落奇点
水管工是离AGI奇点最远的那一群人
123olp retweeted
目前看到最快的,2100 tok/s,赛博新时代的军备竞赛了,很开心是中国的开源模型。这是开源开放的胜利,很快会形成护城河。
太期待了,如果有一天生成速度接近实时,会发生什么🤔
2,100+ tokens/sec per user on DeepSeek V4.1 Flash, at full model quality, on 8× AMD Instinct MI355X. 🚀
Congrats to @lithos_ai on pushing the frontier of inference with persistent megakernels on AMD. Open hardware + instruction-level control = room to innovate.
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding
> 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages.
> 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language.
> 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ 🧵


























