david (@nozamd) header
david (@nozamd) profile picture

computers, AI, photography working on 3D printing slicers, ISBNs and... stuff playing with satellite data @ EUSPA - prev United Nations, Citymapper

a
Joined September 2015
Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)Photo by david (@nozamd)
I don't think they realize how fucked they are lol
Going to check out this threads thing
7
See innocent banter between two sandbox company on X is fun and cute, but if two founders start throwing shit on a random Saturday it stops looking good 👍
16
Am I the only one that thinks it's a bit suspicious that the same day a mathematician working on a human GRH proof (to be published in ~1 year) was meeting with some colleagues to review some of their work, the Anthropic engineer decided to ask Dario to publish this?
卧槽????黎曼猜想可能真的被Claude证明了??? 刚刚,Anthropic的数学家Levent Alpöge在个人网站上发布了一份声明,声称他和Claude一起找到了黎曼猜想(Riemann Hypothesis)的证明。 更离谱的是,他说这套方法还可以直接推广到广义黎曼猜想(GRH),而且他们已经用Lean完成了形式化证明,并公开了相关文件。 要知道,这可是1859年提出、困扰数学界167年的黎曼猜想,数学界最著名的七大千禧年难题之一。 而且Levent Alpöge也不是什么民间数学爱好者。他是普林斯顿数学博士、哈佛前Junior Fellow,现在就在Anthropic工作。 他在声明里还提到了一个非常恐怖的细节: 在证明的最后阶段,Claude的进展速度已经远远超过了他本人。 他甚至形容,Claude在终点线前飞了起来、跳起了舞。 整个证明极其庞大,他尝试生成的完整文字版本甚至超过了25万页,目前先公开了Lean形式化文件。 当然,现在还只是作者本人宣布证明成功,尚未经过数学界独立验证。 但如果这个证明最终成立…… 那我们今天可能真的见证了数学史和AI发展史上最重要的时刻之一。
2
140
david retweeted
Why is this on by default @OfficialLoganK
1
2
103
I'm trying @boxrouter_ai and I asked my agent to launch 32 sandboxes on the cheapest provider. I wonder which provider it chose...
2
8
309
david retweeted
shooting my shot 15 in london i'm an agent native designer, becoming the best (i have 10+ agents running 24/7), i can ship anything to code. always improving, always learning (you can see it in the ads below actually, even though its only been a few days you can tell which ones are older) for example: i can turn your branding into a post generator for whenever you launch something new so you can create assets fast. i love making things look good, if your site doesn't get inbound compliments i won't stop until it does :) i'm high agency, scaled my 3d art agency to 6 figure arr and do 6 figures yearly from web design, however I think now its time for me to do something bigger. previously did design eng at okara(dot)ai, 120k+ users, and at menufit(dot)com, top 10 health and fitness, just under 2m mrr dms are open, for work, collabs, or just to chat!
The ▲ Creative Studio, Website and Product Design teams are growing super fast! We are looking for humans with extraordinary abilities to join us to build the future of the agentic web.
4
2
23
1,214
They are saying it openly now. No more pretence.
Senior Netanyahu govt. officials told @Nadav_Eyal: “We don’t want reconstruction. We don’t want a withdrawal. We’ll leave Hamas as it is and control half the Strip.” Gaza’s physical landscape reflects this, as I reported from a recent Gaza military embed youtu.be/716J29IZAbI?si=htIg…
49
93
11,523
Anthropic's PR circus just forced a real mathematician to dump 12 years of human work unfinished. Hours after their Claude Riemann Hypothesis stunt, AMS President Ravi Vakil hurriedly published a 940-page, ZERO AI proof of GRH by Alessandro Masullo. I opened both PDFs. The contrast is disgusting. Anthropic's Levent Alpöge gave the world a 4-page note and 14 tarballs of machine-generated Lean. In his own PDF he writes that turning it into prose hit 250,000+ pages, calling it "some massive unreadable travesty of 'slop'". He admits he hasn't mastered the math around it and is paraphrasing Claude's slop summaries that "might make zero sense". Sunday night he messaged Dario, and by Friday they blew up the internet with it. That same Sunday, Masullo and Stanford's Brian Conrad spent 4.5 hours at a blackboard at IAS walking Peter Sarnak through Masullo's proof. Conrad has been checking it closely for months, and Vakil wrote it's "so far looking good". They planned to bring Masullo back to Stanford for 1 year to finish polishing the writing. Instead, Vakil had to throw the raw draft online mid-meeting because Anthropic's fucking hype machine couldn't wait. Neither proof is verified yet, and either could still break under review. Even so, read Masullo's closing line on page 6 of his PDF: "If any message at all this work is meant to send to the mathematician who feels discouraged by the pace and tone of the current climate, it is this: we are still here. We are very much alive. We are not going anywhere." Anthropic dumped 250,000 pages of unreadable AI bullshit for attention. Masullo spent 12 years at a blackboard doing real math, and he didn't even get to finish the intro.
126
528
69
4,090
264,016
prompts I never thought I would send
2
40
Why is this on by default @OfficialLoganK
1
2
103
david retweeted
BREAKING: Jensen Huang reportedly painted a fake tunnel on a brick wall, tricking an employee into running headfirst into it.
362
1,570
152
28,001
504,288
time to burn
We’re rolling out a change to increase cache hit rate on Neon by around 4% on Antigravity by massive increasing the cache pool size. This will increase what you can get done in quota by at least 20%. Will issue a reset after it’s done since it burned quota faster!
1
55
Once more, having to switch to GLM 5.3 because Astra and Neon are both idiots
46
Tech leaders are fearmongering about returning to the pre-AI era of... 2022? @jonstewart
71
895
92
6,812
694,557
La comparación con Julio de 2023 es irrelevante. Es un mes en el que la gente coge vacaciones y viaja. La relevante es con noviembre de 2019. Las solicitudes han crecido casi un 12%. En aquellas elecciones la participación fue del 70% justo. Así que sí, se confirma que la participación puede superar el 75% en estas.
🔴 ÚLTIMA HORA | Admitidas 1.099.687 solicitudes de voto por correo para el 29-N: un 58,07% menos que en julio de 2023 cadenaser.com/nacional/2026/…
Readers added context they thought people might want to know
Correos ha admitido 1.099.687 solicitudes. Es un 58,07% menos que en julio de 2023, cuando las elecciones se celebraron en verano con un volumen extraordinario de voto por correo, pero un 11,47% más que en noviembre de 2019. correos.com/sala-prensa/co… abc.es/espana/solicit…
86
949
29
3,489
201,620
I wanna try the Vibe desktop app so bad but I'm out of usage 😭
Just installed Mistral Vibe will try to spend some time over the weekend getting familiar with @MistralAI First feedback I have is just having a default workspace would be great - like deepseek harness, minimax code and others so that from the start I don't have to create any dedicated project folders, this would be good for all the things that do not belong to a specific project, one off things 😎 other than that, app looks clean follows the same look we know from minimax code, qoder and others - which is good imo, clean and simple. more feedback to come in the following days, will be testing the Code and Work options of the Vibe app, will try to show a bit more of how this all works
1
2
83
so T3 is now paying Theo to talk good about it wow
Latest stable release of T3 Code should be a significant improvement in Antigravity/Gemini support btw 👀
3
121
Jumping back into open-source on the side! Introducing REPLico, an easy way to build MCP servers like Codex computer-use, where the agent calls tools by running code in a persistent Node Repl environment. This allows the model to chain actions, filter outputs, and verify completion within a single tool call, reducing round-trips and improving token efficiency. REPLico lets you easily get the same behavior (~as fast or faster than `cua_repl`!) without having to set up Node and behaviors like state persistence yourself. You might be surprised to hear I DON'T recommend this to be used by the vast majority of MCP servers! Most MCP servers should follow the spec, and allow agent harnesses like Codex to implement features like code-mode internally. However, this does make sense for some specific use cases (like computer-use agents). It's super easy to get started with ReplicoKit. I was inspired by FoundationModels' API: define your `Tool`s in Swift using the `ToolArguments`/`Arguments` macros, then conform your app's entry point to `ReplicoApplication`. github.com/julianschiavo/REP…
5
1
14
1,034
what a tweet
Replying to @willcb
Neon is based on Hydrogen, as well as Carbon. Carbon is based on Argon, which itself was based the unreleased Helium model. Sodium is inspired by Neon and Helium but was trained concurrently.
1
94