@wateriscoding

orchestrator and diff reviewer

Joined November 2023
Finally wrote a blog mentioning some of my resources... They are not much but i'll keep updating it (maybe). Also, my first blog so suggestions are welcome... waterwater.bearblog.dev/my-m…
37
45
2
599
56,807
I believe him
JUST IN: Professor Jiang claims AI isn’t real intelligence & is instead “manipulated by humans somewhere in India.”
Readers added context they thought people might want to know
Jiang Xueqin styles himself "Professor Jiang" but is a high school teacher with a Yale BA in English literature and no university faculty position. His claims misrepresent AI as live human manipulation rather than output from trained neural networks. en.wikipedia.org/wiki/Jiang_Xue… indiatoday.in/world/story/pr… en.wikipedia.org/wiki/Large_lan…
1
1
9
560
water retweeted
how it feels not using jev, instinct or wispr flow in this economy
26
32
6
891
18,004
opened x after a month, why are all the people here suddenly talking about jav now? and why is everyone talking like it popped up recently, i remember it being around for years now
12
318
Almost a year of building in my room. Meet Sudo. A little desk robot that talks, plays chess, and gets way too excited when it beats me. Your AI. Your personality. Your mods. 25 Founder's Edition units. Preorders live: presencelabs.in
28
22
4
123
8,789
See how agents work. Then build it. Skillmaxing two sequenced paths for engineers: Python with AI → FastAPI, Agents SDK, MCP, tracing, deploy System Design → notes, company cases, from-scratch builds 10 lessons free. Then ₹350 for the first month skillmaxing.xyz/
2
14
64,003
Learning microservices (and probably exploring pi code if my headache stops) this weekend. What are you guys doing!? Any resources on both of them are welcome 🤗
2
4
431
Found an actual user of my product and that made my day ✨ Such a shame this buried under the algo. What does even with these days.
hahahaha actually clutched last moment with the limits and created this - Rabbithole 🕳️ start reading the topic -> explore more related topics -> create the trail -> share if interesting created in like ~4 hours 😅 made for phones(primarily), only wikipedia source for now, might expand to other sources later, if i'm interested in this 😅
1
11
767
Good morning
38
4,263
41
47,498
433,262
water retweeted
BetterShot v0.4.0 is out. This one is a full rebuild. The recording pipeline, the floating bar, and both editors were rewritten from scratch. The work that used to send you to After Effects now lives in the app. What's new: • Auto zoom. Your cursor position is tracked while you record. Clicks turn into zoom cues, and the camera moves the way a screencast editor would. Every cue is editable on the timeline. • A cursor the editor draws. The pointer is no longer baked into the video. Its path, clicks and shapes are saved next to the recording, so you can restyle it afterwards. A spring smooths a shaky hand and lands clicks exactly on the spot. • Face cam. A round camera bubble you can drag anywhere while recording, and drag again in the editor. Works with Continuity Camera too. • Microphone support. Narration records as its own audio track, kept separate from system audio. There is a live level on the bar, and muting writes nothing at all instead of writing silence. • A new editor. Six tabbed inspectors instead of one long sidebar. Multi clip timeline, per clip speed, transitions, split screen, 3D tilt, color grading, captions, keyboard overlays, and blur masks that are destroyed in the export rather than covered. • Cloud sharing. Files compress locally before anything uploads, then publish straight to your own Cloudflare R2 bucket. No proxy, no third party service, keys in the Keychain. Your files stay yours. Also: crash resilient recording, multi display capture, and single window recording. Free, for macOS. Link in the comments:
7
2
40
3,477
funny how this used to be a trauma a few years ago, and now people are making jokes and games around it
i made a game where you literally have to find a missing semicolon in a codebase of 187,421 lines the catch is the entire codebase is printed out and taped to the walls of an office and you have to inspect it line by line Play it here: semicolon.trm.sh/
9
760
hahahaha actually clutched last moment with the limits and created this - Rabbithole 🕳️ start reading the topic -> explore more related topics -> create the trail -> share if interesting created in like ~4 hours 😅 made for phones(primarily), only wikipedia source for now, might expand to other sources later, if i'm interested in this 😅
i have until tomorrow afternoon before my plan ends. what should i build??
3
2
1
13
1,605
926
18,323
1,880
188,623
5,732,021
Where is the real Tibo? The Tibo OpenAI put on camera yesterday barely resembles the guy in his profile pic. I think sometime in 2024 Sam replaced him with THIBO-3000 and only rolls him out occasionally so nobody gets suspicious. I think they already achieved AGI, and this interview confirms it.
170
21
17
846
116,074
Yeah... Pretty sure this counts as blasphemy
才发现 Codex「额度用完还能跑完任务」的福利前几天还有的,结果被两位傻逼搞没了🤬 时间线 👇 8/20:OpenAI 员工 Angel 发帖:额度打到 100%, 也不会掐断正在跑的任务,是 OpenAI 刻意设计的,“在别人干活干到一半时停掉,没什么帮助”。这时福利还在 8/21:网红 Theo 贴脸开大,转发并当场示范怎么滥用:开 /goal 跑长任务,额度归 0% 之后,又烧掉至少 $800 的 tokens 同一天,Theo 智商 250 的实习生 Maria 直接把这招做成了一个叫 Endless 的 harness。 原理是这样的: - 开一个 turn,给 agent 塞个 wait_for_user_input 工具 - 任务做完不结束 turn,而是调用这个工具等待 - 你的下一句话当工具结果喂回去 整个会话永远锁在同一个 turn 里,quota 检查永远轮不到触发 宣传语原话:「infinite inference (kinda)」,榨干你订阅的每一滴 8/22:Endless 帖子爆了,一堆人拿它在归零后继续白嫖,眼看演变成大规模滥用 Maria 当天道歉:“没想到这个小点子危害这么大,如果导致大规模滥用我愿担责” 当场下架 npm 包、关掉网站、删了 GitHub 仓库 最骚的还在后面: 删完库,她转头喊话 Codex 团队:“希望你们排查一下,把这些账号封了,这样我们就安全了 😊” 不是,你傻逼吧,居然呼吁 Codex 团队封掉用你工具的账户???哪有人那么背刺信任你的人的 💩 这一轮风波的结果就是:现在 Codex 额度用完,任务立马停 反观: Theo 悄悄删了那条 $800 示范帖 Maria 的 Endless 下了,帖子也删了 点火的全身而退,火上加油的道了歉,最惨的是连 Endless 都没听过的我们 我们的 「额度用完还能跑完任务」福利是 TMD 真的没了啊,气死了 😤 nitter.cf/angelbrodin/status/209…
1
4
511
So, this is the guy landing the resets ☝🏻 Can clearly see the pressure of handling people's cries and complaints on resets and limits in his eyes🫩
My full interview with Tibo (@thsottiaux) 0:45 Tibo's Lessons from Google DeepMind 4:22 Building OpenAI’s Relentless Culture 7:23 Astra & Next Gen Models 11:18 How Fast AI Changes Developer Workflows 14:27 ChatGPT & Codex Merging 20:25 OpenAI vs. Anthropic 23:37 Why OpenAI Keeps Resetting Limits 30:25 Recursive Self-Improvement 32:00 Dangers That Caused "The Pause" 34:13 Will Ultra Fast Become the Default? 43:20 Why Everyone Needs to Try AI
4
413
Everything i am reading in a single place - 9 #1 Netflix has put out a detailed paper on how they are using LLM-as-judge in their recommender systems. A neat research log of how they treat an LLM judge like a lifelong agent going through various phases over just one dedicated role for a turn/attribute. arxiv.org/abs/2608.18300
5
6
1
105
4,040
some ideas look very good on paper and they should remain on paper
2
1
4
25
7,284
i have until tomorrow afternoon before my plan ends. what should i build??
1
1
19
1,868
Successfully fumbled another social interaction today ✅
4
27
775