liquan
@flyq_crypto
Followers
357
Following
5K
Media
52
Statuses
706
Building state proof, block witness. #INTJ Writeups(https://t.co/19iqmMOxPT) about #sumcheck, #ipa, #folding, #kzg, #verkle
Field
Joined October 2021
GPT-5.1 is out! It's a nice upgrade. I particularly like the improvements in instruction following, and the adaptive thinking. The intelligence and style improvements are good too.
2K
2K
14K
ZK 真的是寒冬了，Jim 之前发的那篇在binary field 提高zk效率的paper，V 神都专门写了一篇blog
Irreducible's time has come to an end. Long story short, @radi_cojbasic and I ultimately came to the realization that we couldn't sustainably build the kind of deep tech business in ZK that motivated us.
0
0
0
This industry is rife with scalpers, and everyone's gotten used to it.
Any megaETH sale participant who locked their tokens for a year is required to (i) acquire the tokens for their own account without any resale or transfer intention, and (ii) refrain from any transfer, resale, or hedging transaction that would violate applicable laws. Anybody
0
0
0
让 cluade 优化掉一个新申明的 bool 变量，在使用这个 bool 做判断的地方，它使用看起来毫不相干的几个变量构造了一个判别式来代替。我第一反应它在乱改，直接说 are you kidding me，它乖乖的改回来，仍然需要申明一个变量。后面我仔细想了一下，发现它第一次修改是对的。ai 是对的，我错了
0
0
1
V神很多哲学性的Blog我都有点看不懂，但是这么一翻译，get！
这篇文章写的确实很好，借助 AI 帮忙解读一下（解读的很好，比直接翻译更容易看明白） 哈啰，大家好。 你是不是也经常在网上看到一些争论，感觉某一方说得天花乱坠、旁征博引，听起来“非常智能”，但你心里总有个疙瘩，觉得“哪儿不对劲”？
0
0
1
eth: ICO = 49% Team + Foundation = 10% Mining = 41% mega: ICO + VC = 30% Team + Foundation = 17% Mining = 53% Make ICO Great Again
有没有人对比 mega 的 token 分配和 eth 的分配？
1
0
1
In the spirit of transparency, Here is the list of sale participants that we believe represent our core community over the past couple of years. They represent 9.5% of addresses and 25% of USD that received allocation in the sale:
375
54
742
有没有人对比 mega 的 token 分配和 eth 的分配？
Over half of total supply is reserved for KPI-based distribution Conditions → Gated to $MEGA stakers → Time staked impacts rewards (10d min, 30d max) → Unvested token can stake, but rewards share lock terms This puts rewards in the hands of those aligned with network growth
0
0
1
我觉得这个对上手新项目很有用，哪天试一下
Cognition Labs 推出 CodeMaps：全新的代码表示方式 软件工程中有一个普遍痛点：理解复杂代码库很难！这不仅仅是新手工程师的困扰，即使资深开发者也常常花费大量时间在代码引用导航跳转上。根据 Stripe 研究，遗留代码维护已成为生产力最大的拖累，而现有 AI
1
0
1
哈哈哈，能够降低开会时走神的负罪感，挺好的理论。 还有一条，信息的筛选，如果潜意识里面觉得不重要，那些信息就被过滤掉了，如果觉得重要，就听得进
或许是因为adhd的大脑接受信息的带宽非常强，需要多种信息刺激， 而开会-语言输出的信息量实在是有限。 ADHD有一个"猎人基因假说解释",adhder其实是古代猎人后代，其大脑天然适应的是野外的环境:
1
0
1
以后机器人的通用语言是中文
Smoking gun: Pretty sure Cursor’s new Composer-1 is a fine-tuned Chinese model. As I was building, it switched its inner monologue to Chinese, and I can't get it back to english. @simonw
0
0
0
最近的经验，claude + sonnet 4.5 感觉比 codex + gpt5-codex-high 效果要好
GPT-5 is the best, I use Cursor Pro, and my previous model was primarily Gemini Pro 2.5. I recently ran into a tricky problem that kept failing. Gemini Pro 2.5 proved incredibly powerful, helping me understand the relevant reference code, but it failed to fix my own erroneous
1
0
1
学习了
今天看到一篇文章，作者（Joe Mag）是亚马逊的一位工程师，他所在的一个资深工程师团队，过去三个月里借助 Coding Agent，代码产出效率是普通高效团队的 10 倍。 但作者可不像傅老板那样说人类程序员要被替代，毕竟人家是专业程序员，他反而是来浇冷水的。他用亲身经历告诉我们：当你的速度提升 10
0
0
0