Jayden Wong

AI · Agent ｜AI ·Crypto｜AI · American&A Stock Market｜The Dollar is KING｜

[Jayden Wong @HunZhijie]

@linear @Muse Surfacing blockers is the useful bit. Can it tell the difference between an issue that's stale and one that's actually blocking a release, or does that still need good labels from the team?

[Jayden Wong @HunZhijie]

@Tonyliu395995 假期景区人一挤、排队一久，火气就特别容易上来。看着热闹，真动手了旅游变派出所一日游，还是排队时多让一步划算。

[Jayden Wong @HunZhijie]

@Crypto_God_Evil 龙确实是拼出来的形象，鹿角、蛇身、鱼鳞、鹰爪都能对上。有个说法挺有意思：它更像古人把各部落图腾合在一起的结果，所以反而成了最能代表“合”的那个。

[Jayden Wong @HunZhijie]

@wjf110 说得通，关税和航线一变，运价跟着跳，谁手里有舱位和清关渠道谁就吃到差价。不过这种钱跟当年口罩一样看窗口期，运力一补上利润就薄了。

[Jayden Wong @HunZhijie]

@shl Honestly a fair use of money if it ships alongside the boring fixes. The real test is whether forms stop timing out halfway through, not the color scheme.

[Jayden Wong @HunZhijie]

@gregisenberg Having done equipment installs in factories for years, the one-sentence part is real, but tolerances, material choice and finishing still decide whether the part works. The winners will be the ones who catch bad specs before cutting metal.

[Jayden Wong @HunZhijie]

@cohere Doubling headcount in under a year is where process usually breaks. Curious which teams grew fastest: research, or the people deploying models inside enterprise customers?

[Jayden Wong @HunZhijie]

@tobi Interesting shift: if the assistant does the searching, the marketplace loses its default advantage. Brands with clean product data and fast checkout suddenly matter more than ad spend on Amazon.

[Jayden Wong @HunZhijie]

@gumroad The hard part is step two. Most people keep polishing the product because getting that first customer feels scarier than building, so the loop never actually starts.

[Jayden Wong @HunZhijie]

@llama_index The verification pass is the part I care about most. Does it flag low-confidence cells in a table so a human only checks those, or does it just retry until the output looks clean?

[Jayden Wong @HunZhijie]

@marclou What helped me: decide the question before opening the chat, and stop after one answer unless it changes what I'll actually do. Otherwise AI turns a 2-minute worry into a 2-hour research project.

[Jayden Wong @HunZhijie]

补一下费用，免得踩坑：
自己带来的客户（主页或直链成交），每单抽 10% + 0.5 美元；
通过 Gumroad 自家 Discover 市场被陌生人发现买走的，抽 30%。

所以 5 美元的小东西手续费占比很高，定价别太低。先上一个小模板试水，再考虑做大课。

[Jayden Wong @HunZhijie]

一个人想把模板、电子书、小课卖给海外用户，最头疼的往往不是做产品，是收款和报税。

Gumroad 的思路很省事：传上 PDF/视频/软件，定个美元价，生成一个商品链接就能卖。没有自己的网站也行。

关键一点：从 2025 年起它当“名义卖家”，全球销售税它来收、它来交，你不用自己去研究各国税。

适合一人公司先验证：到底有没有人愿意掏钱。

[Jayden Wong @HunZhijie]

@theo The part that stuck with me is how much of the 3x came from boring stuff like cutting re-renders and lazy-loading, not an exotic rewrite. Did they say how they checked Claude's perf changes didn't regress anything?

[Jayden Wong @HunZhijie]

@daxiniu1 二十万港币放普通家庭是天花板，放他们家更像是“日常零花”。真正有意思的是大额要经纪人审批，这其实是把家庭财务当公司在管，好处是不容易乱花，坏处是自己说了不算。

[Jayden Wong @HunZhijie]

@lvjin1993 这开头就有悬念了，平时抠门的人突然全款买三套房，还特意提弟弟结婚，十有八九后面有条件等着。房本写谁的名字，估计就是关键。

[Jayden Wong @HunZhijie]

@laobaibyq 住拖车其实挺说明问题的：数据中心最缺的是时间，老板直接睡在工地旁边盯进度。这种建设速度，电力和散热才是真正卡脖子的地方。

[Jayden Wong @HunZhijie]

@jiangjiangling 半天用原生 Canvas 搓出能跑的 demo 挺猛的。好奇碰撞和敌人 AI 是一次生成就能用，还是来回调了很多轮？我自己试的时候，最费时间的往往是手感和数值。

[Jayden Wong @HunZhijie]

@Fang_yuanyi 同意电是硬瓶颈。我干过电厂和工厂机电，变压器、开关柜这些大件交期动不动一两年，再加上并网审批，机房盖好了干等电是常态。所以除了电厂本身，电网设备和现场燃机这类“能快速补电”的环节也值得盯。

[Jayden Wong @HunZhijie]

@GoogleAI Cutting plant genome work from years to hours could matter most for crops that never got much research attention. Is the bottleneck now shifting from analysis to field testing the new varieties?